Histological Hyperspectral Breast Cancer Recurrence Database (HistologyHSI-BC Recurrence)

Histological Hyperspectral Breast Cancer Recurrence Database (HistologyHSI-BC Recurrence) | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Histological Hyperspectral Breast Cancer Recurrence Database (HistologyHSI-BC Recurrence) Laura Quintana-Quintana, Esther Sauras-Colón, Alessio Fiorin, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7242335/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Metastasis occurs in nearly 1 out of 3 breast cancer (BC) patients and significantly reduces survival rates, particularly in cases of distant metastases. As most distant metastases develop after diagnosis (i.e., recurrence) and remain incurable, there is a critical need for prognostic biomarkers to assess recurrence risk. Multimodal data analysis has emerged as a promising approach to integrate diverse information, offering a more comprehensive perspective. This study introduces the Histology HSI-BC (hyperspectral imaging - breast cancer) Recurrence Database, the first publicly accessible multimodal database designed to advance BC distant recurrence prediction. The database comprises 47 histopathological whole-slide images, 677 hyperspectral (HS) images, and clinical and demographic data from 47 BC patients, of whom 22 (47%) experienced distant recurrence over a 12-year follow-up. Histopathological slides were digitized using a whole-slide scanner and annotated by expert pathologists, while HS images were acquired with an HS camera coupled to a bright-field microscope. This database provides a promising resource for studying BC recurrence prediction and personalized treatment strategies by integrating the aforementioned multimodal data. Biotechnology and Bioengineering Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Background & Summary In 2022, breast cancer (BC) was the most common type of cancer in women, with an incidence of 23.8%, and the leading cause of cancer-related death among women, accounting for 15.4% of all cancer-related deaths 1 . Cancer cells can spread from the primary tumor to other parts of the body, which is known as metastasis and is the main cause of death in most cancers 2,3 . Metastasis occurs in nearly 1 out of 3 patients diagnosed with BC and can appear in the axillary lymph nodes (regional metastasis) or in other organs (distant metastasis) 4 . Overall, the 5-year survival rate after diagnosis of BC is 91%. However, this rate is higher in patients with tumors located exclusively in the breast (99%) than in patients with regional metastasis (86%) or with distant metastasis (31%), where survival decreases dramatically 4 . Women with distant metastases may have either de novo distant spread, where distant metastases are already present at the time of diagnosis, or develop distant metastases after an initial diagnosis and treatment, which is known as recurrence 5 . While de novo cases account for approximately 25% of metastatic BC diagnoses, the majority result from recurrence 6 . Whether de novo or recurrent, distant metastases remain incurable 7,8 . Certain classic prognostic factors are associated with the risk of developing distant metastasis, such as age , tumor diameter , stage , tumor grade , tumor type or lymphovascular invasion (LVI) 8 . Additionally, studies have identified other biomarkers with prognostic value in the disease that may be associated with metastasis, including genetic alterations, circulating tumor cells and circulating tumor DNA, biomarkers of response to immunotherapy and gene expression platforms to predict the risk of recurrence 9–11 . However, to date there is no consensus for the implementation of most of these biomarkers in routine clinical practice. Therefore, there continues to be a growing interest in identifying specific prognostic biomarkers that allow determining the probability of developing metastasis. Cancer detection relies heavily on imaging methods like X-ray, ultrasound, and magnetic resonance imaging 12 . However, treatment decisions require a conclusive histopathological diagnosis, which is obtained from a tissue biopsy. BC can be broadly categorized into in situ carcinoma and invasive carcinoma. Among these, ductal carcinoma in situ (DCIS) represents the most prevalent subtype of in situ carcinoma, while invasive ductal carcinoma (IDC) is the most common subtype of invasive carcinoma. Nevertheless, given the heterogeneity of BC, the accurate identification of these subtypes among other histological subtypes requires extensive expertise and a deep understanding of breast pathology 13 . The rise of digital pathology, which leverages whole-slide images (WSIs), has revolutionized research and diagnosis in pathology, particularly in cancer, by enabling more efficient data sharing across institutions and promoting remote collaborations. WSIs are high-resolution digital images of traditional glass pathology slides, which can be viewed, analyzed, and shared on a computer screen 14,15 . The use of WSIs also paves the way for computational pathology, which started from the use of traditional image analysis methods to advanced machine learning (ML) and deep learning (DL) algorithms 16,17 . Remarkably, these novel approaches offer the potential to integrate multiple data modalities, extending beyond histopathology image analysis. This includes linking histopathological images with clinical factors, such as prognosis and genetic mutations, thereby enhancing BC diagnostics 18–20 . Beyond conventional methods, other imaging modalities show promising potential for improving the diagnosis and prognosis of BC patients. Among these, hyperspectral (HS) imaging (HSI), combines traditional imaging with spectroscopy to capture both spatial and spectral information. Each material interacts uniquely with emitted radiation, reflecting and absorbing it in a way that creates a distinct radiance vector, often named spectral signature . HSI sensors can capture these spectral signatures, acquiring significantly more data than standard RGB (Red, Green, Blue) cameras and extending imaging capabilities beyond human vision (e.g., near-infrared (NIR) HS sensors can capture wavelengths ranging from 900 to 1,700 nm) 21 . In recent years, the use of HSI in medicine has begun to achieve promising results regarding cancer detection by utilizing cutting-edge ML algorithms to process the high amount of HS data 22–24 . In the existing literature for medical histological applications, HSI has been used to identify pancreatic neoplasms with different prognoses 25 , quantify Ki67 as a prognostic factor in lymphomas 26 and study the interactions between tumor cells and immune cells of the tumor microenvironment in response to immunotherapy in lung cancer 27 , obtaining promising results. Research in this area is still in its early stages, and the number of published studies remains limited. Regarding WSI and clinical and demographic databases, The Cancer Genome Atlas (TCGA) is one of the main publicly available sources for hematoxylin and eosin (H&E)-stained WSIs and associated clinical and demographic data 28 . The main challenge of this database is the lack of annotations, which makes the subsequent analysis of these WSIs difficult. The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) is another database that provides clinical, demographic, and molecular data of over 2,000 BC cases 29 . However, it primarily focuses on genomic and transcriptomic data and does not include WSIs. Several publicly available databases contain H&E-stained WSIs 30 , but most include annotations focused on specific cell types in BC, such as tumor-infiltrating lymphocytes and inflammatory cells. We have identified one database - Breast Cancer Semantic Segmentation (BCSS) that provides specific annotations of tissue compartments, distinguishing between tumor and healthy tissue 31 . Regarding medical HS databases only two major sets were identified: in 2022, Zhang et al. 32 introduced a large-scale database for HS microscopic images of precancerous lesions in gastric cancer, and in 2024, Ortega et al. released the HistologyHSI-GB dataset 33 , focused on HS glioblastoma histology. None of those databases included clinical or demographic data, this being one of the main challenges in this field. There is a limited availability of comprehensive, high-quality databases, which hinders the broader application of clinical and demographic data, WSIs and HSI in clinical practice and research. The lack of such databases makes it difficult to fully explore the potential of HSI together with conventional practices for diagnosing and predicting disease outcomes, such as recurrence in BC. This paper presents a publicly accessible database designed to investigate specific prognostic biomarkers for predicting the likelihood of BC recurrence due to distant metastasis. The HistologyHSI-BC Recurrence Database includes clinical and demographic data from BC patients, along with WSIs and HS images obtained from their primary tumor samples. This database is intended to evaluate the ability to predict recurrence due to distant metastasis over a 12-year follow-up period. Biopsies from 47 patients diagnosed with BC were extracted, sliced and stained with H&E, 47 WSIs and 677 microscopic HS images were taken, and their clinical and demographic data were collected. Among these patients, 22 experienced distant recurrence. A schematic overview of the study workflow is presented in Fig. 1. The HistologyHSI-BC Recurrence Database will benefit researchers by providing a comprehensive, multimodal database that integrates WSIs, HS images, and clinical and demographic data from BC patients. This resource enables the development and refinement of predictive models for BC recurrence due to distant metastasis, starting to fill a significant gap in available databases. Researchers can leverage this data to explore innovative ML approaches, enhance diagnostic accuracy, and identify novel biomarkers for BC recurrence. Additionally, the database promotes reproducibility, facilitates collaboration across institutions, and accelerates research in computational pathology, potentially improving personalized treatment strategies and benefiting broader cancer research efforts. Methods Patients Selection, Eligibility Criteria and Ethics Approval This is a retrospective case-control study carried out on 47 BC patients diagnosed with IDC, now called invasive breast carcinoma of no special type 34 , between 2006 and 2015. Cases include 22 patients who experienced recurrence due to distant metastasis during the 12 years following diagnosis. The remaining 25 patients who did not experience recurrence during the 12 years of follow-up are included as control group. Table I. Eligibility criteria for patient inclusion. Inclusion Exclusion A diagnosis of IDC Receipt of neoadjuvant treatment, as it is known to modify the tumor microenvironment Representativeness of IDC tissue in surgical biopsy Recurrence occurring in the breast rather than in distant organs A clinical history with complete clinical and pathological data Presence of distant metastases at the time of diagnosis Patient’s agreement to be included in the study. The exclusion criteria included Failure to meet any of the inclusion criteria The study was approved by the Drug Research Ethics Committee of the Institut d’Investigació Sanitària Pere Virgili (IISPV), Tarragona, Spain, under reference number 134/2022. The samples used in this study were obtained from Biobank IISPV-Node Tortosa, Tarragona, Spain, following the principles of ethical conduct and data protection. All participants whose samples were stored in the biobank have previously signed an informed consent form, explicitly authorizing the collection, storage, and future use of their biological materials and associated data for research purposes. The processing, communication and transfer of personal data of all participants comply with the provisions of Organic Law 3/2018, of December 5, on the Protection of Personal Data and Guarantee of Digital Rights and with Regulation (EU) 2016/679 of the European Parliament and of the Council, of April 27, 2016, on the protection of natural persons with regard to the processing of personal data and the free circulation of these data, and repealing Directive 95/46/EC (General Data Protection Regulation). The data generated and collected during this study are anonymized to ensure the security of the information, safeguarding the confidentiality and privacy of the patients. Data Collection Clinical and Demographic Data The data collection process involved extracting information from clinical records, including demographic and clinical data, which were following Table II. Table II. Description of the study variables. Attribute Definition Format Demographic Data Sex Patients' gender 1: Female Race Patients’ race 1: White Ethnicity Patients’ ethnicity 1: Hispanic Menopausal status Menopausal status of the patient 0: Premenopause 1: Postmenopause Diagnosis Dx surgery Type of surgery 0: Mastectomy 1: Lumpectomy Dx age Difference between the date of diagnosis and the date of birth Years Tumor Data Tumor diameter Maximum diameter of the irregular shaped tumor Millimeters Tumor histologic grade Degree of differentiation of tumor cells, reflecting how different they are from normal breast cells 1: Grade 1 2: Grade 2 3: Grade 3 L ymphovascular Invasion (L VI ) Presence of tumor cells within lymphatic or blood vessels 0: Negative 1: Positive P erineural Invasion (P NI ) Ability of cancer cells to proliferate around peripheral nerves and, eventually, invade them 0: Negative 1: Positive Tumor Stage T (tumor) Tumor size assessed by pathological evaluation 1: T1 2: T2 3: T3 4: T4 N (node) The cancer has spread to the lymph nodes assessed by pathological evaluation 0: N0 1: N1 2: N2 3: N3 M (metastasis) Status of breast cancer spreading to a different part of the body 0: M0 Tumor Molecular Markers (determined by IHC) ER Status of estrogen receptors (ER) 0: Positive ER for < 1% of the TC 1: Positive ER for ≥ 1% of the TC PR Status of progesterone receptors (PR) 0: Positive PR for < 1% of the TC 1: Positive PR for ≥ 1% of the TC HER2 Status of human epidermal growth factor receptor 2 (HER2) 0: Negative (IHC = 0/1+, or IHC = 2+ and FISH = negative) 1: Positive (IHC = 3+, or IHC = 2+ and FISH = positive) KI67 Index quantifying KI67 expression to measure how fast cancer cells are dividing in a tumor 0: KI67 index < 20% 1: KI67 index ≥ 20% Molecular subtype Classification according to IHC status of ER, PR, HER2 and Ki67. 0: Luminal A (ER+/PR+, HER2-, KI67-) 1: Luminal B HER2- (ER+/PR+, HER2-, KI67+) 2: Luminal B HER2+ (ER+/PR+, HER2+, KI67-/+) 3: HER2+ (ER-, PR-, HER2+, KI67-/+) 4: Triple negative (ER-, PR-, HER2-, KI67-/+) Lymph Nodes Status LN status Status of the spreading of tumor cells to the lymph nodes, including sentinel and non-sentinel lymph nodes 0: Negative 1: Isolated TC (CC ≤ 0.2 mm) 2: Micrometastasis (0.2 mm 2 mm) LN ITC number Lymph nodes with isolated tumor cells Natural number LN MICRO number Lymph nodes with micrometastasis Natural number LN MACRO number Lymph nodes with macrometastasis Natural number LN number Lymph nodes removed during sentinel lymph node biopsy and/or lymph node dissection Natural number SLN number Lymph nodes removed during sentinel lymph node biopsy Natural number SLN status Presence (or absence) of tumor cells in the sentinel lymph node(s), which are the first lymph node(s) where cancer cells might spread from a tumor. 0: Negative 1: Isolated TC (CC ≤ 0.2 mm) 2: Micrometastasis (0.2 mm 2 mm) Tumor Treatmen t Tx hormonal Patient received (or not) hormonal treatment to stop estrogen and progesterone from attaching to their receptors 0: Not received 1: Received Tx CT Patient received (or not) adjuvant chemotherapy (CT) after the surgery 0: Not received 1: Received Tx trastuzumab Patient received (or not) trastuzumab, a targeted therapy used to treat HER2+ breast cancer (Molecular_subtype = 3) 0: Not received 1: Received Tx RT Patient received (or not) radiotherapy (RT) 0: Not received 1: Received Follow-up Metastasis type Status of cancer spreading from the primary tumor to other organs during the follow-up period 0: No evidence of local or distant metastases 1: Metastasized on nearby tissues or lymph nodes 2: Metastasized in distant organs from primary site 3: Both local and distant metastases are present DFS The time a patient survives without any signs or symptoms of cancer after finishing primary treatment. It is calculated as the difference between the date of relapse and the date of diagnosis. If the patient did not relapse, the date of last follow-up is used instead. Months Vital status 0: Alive 1: Deceased Death cause 0: Other causes / Still alive 1: Cancer OS The time from the date of cancer diagnosis that patients remain alive. It is calculated as the difference between the date of death and the date of diagnosis. If the patient did not die, the date of last follow-up is used instead. Months CC, cluster cells; CT, chemotherapy; DFS, disease-free survival; Dx, diagnosis; ER, estrogen receptors; FISH, fluorescence in situ hybridization; HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry; ITC, isolated tumor cells; KI67, proliferation index; LN, lymph node; LVI, lymphovascular invasion; MACRO, macrometastasis; MICRO, micrometastasis; OS, overall survival; PNI, perineural invasion; PR, progesterone receptors; RT, radiotherapy; SLN, sentinel lymph node; TC, tumor cells; Tx, treatment. Histopathology WSIs Paraffin blocks of primary tumor biopsies with sufficient representative IDC tissue were obtained from the Biobank IISPV-Node Tortosa, Tarragona, Spain. The samples were processed in the Pathology Department, where 2 µm-thick sections were prepared from each paraffin block and stained with H&E according to the instructions of the manufacturer. The slides were sealed with coverslips using dibutylphthalate polystyrene xylene (DPX) mounting medium for subsequent digitization and HS microscopic image acquisition. The H&E-stained slides were digitized with the Pannoramic 250 Flash III WSI scanner (3DHISTECH Ltd., Budapest, Hungary) at 20× magnification (0.2433 µm/pixel) using MRXS image format. WSIs were visualized using QuPath 35 for technical validation and annotation, and anonymized using the SlideMaster software (3DHISTECH Ltd., Budapest, Hungary). The annotation process of each WSI was manually performed by pathologists using diverse colors to distinguish between IDC, healthy tissue, and DCIS. The annotations were made with the following color scheme: IDC was outlined in blue, healthy tissue in green, and DCIS in red (Fig. 2a). Moreover, within each of these three tissue compartment types, different regions of interest (ROIs), surrounded by yellow line, were identified and annotated to subsequently acquire the HS image using the HS microscopic system (Fig. 2a). These ROIs were selected to ensure the inclusion of representative areas of each class (IDC, healthy and DCIS), capturing relevant spectral variability for further analysis. HS Images The HS images were captured using a HS microscopic system (Fig. 2b). The system features the Hyperspec® VNIR (Visible and Near Infrared) A-Series camera (HeadWall Photonics, Fitchburg, MA, USA), a pushbroom HS camera that captures data by scanning the sample spatially. The camera is equipped with a charge-coupled cevice (CCD) sensor that covers a spectral range of 400-1,000 nm, capturing 826 spectral bands across 1,004 spatial pixels per line. It offers high spectral resolution with a slit image full width at half maximum (FWHM) of 2.5 nm and a pixel size of 7.4 μm. Data are acquired with a 12-bit ADC (Analog-to-digital Converter), and each HS line has a size of 1,004 × 826 pixels and requires 1.6 MB per line on disk for storage. The microscope used is the OLYMPUS BX-53 (Olympus, Tokyo, Japan), with LMPLN-IR (5×, 10×) and LCPLN-IR (20×, 50×) objective lenses optimized for infrared imaging. The system uses a 100 W TH4 halogen lamp (Olympus, Tokyo, Japan) as the light source, covering a wavelength range from 400 to 1,800 nm and supporting both transmittance and reflectance light paths. To acquire full HS images, the pushbroom camera requires spatial scanning, which is facilitated by a SCAN 130 × 85 scanning stage (Märzhäuser, Wetzlar, Germany). The stage ensures high precision (±3 μm accuracy) as it moves the sample, keeping it aligned with the objective and light path. Furthermore, an RGB camera, the acA5472-17uc (Basler AG, Ahrensburg, Germany), provides real-time visualization of the sample to navigate it without the need of using the microscope binoculars. It is a 20 MP compact camera with a Sony IMX183 CMOS sensor (Tokyo, Japan), 5,496 × 3,672 resolution, and 17 fps. It features USB 3.0, a C-mount, and supports hardware/software triggers. Calibration of the HS images is necessary to ensure the data accurately represents the sample's spectral signatures. The HS microscope captures spectral signatures for each pixel, but factors like the sensor's response, light transmission, and the light source can affect accuracy. The calibration process involves normalizing the pixel values of the HS image by adjusting them based on a white reference (WR) and a dark reference (DR). WR is obtained by focusing on an empty area of the slide at the same working distance. This ensures no sample material is present, allowing the frame to record the maximum signal the sensor can measure for each pixel and wavelength under the given conditions (e.g., exposure time, light intensity, and slide properties). Conversely, the DR is captured by completely blocking light transmission to the HS camera. This frame captures the minimum signal levels detectable by the sensor for each pixel and band, as well as dark current information from the CCD. Ideally, DR values approach zero; however, higher values may occur due to intrinsic sensor noise. To enhance the reliability of the calibration process, 100 frames are collected for both the WR and DR, ensuring that averaging reduces potential errors. Finally, the calibration of the HS image is achieved using Equation (1), which relates the calibrated HS image (ri ) to the raw HS image (Rawi ). In-house software was developed to serve multiple functions in the HSI acquisition process. It displays the RGB image to facilitate sample navigation under the microscope and ensures synchronization between the HS camera and the scanning platform by aligning their frame rate and platform movement. After capturing the HS image, the software removes the extreme bands from the raw HS image (reducing the spectral range from 400-1,000 nm to 400.5-938 nm), as these bands contain significant noise, and then saves the raw HS image. The calibrated HS image is then generated, using Equation (1), and saved on memory as five-digit 16-bit unsigned integers (uint16), where the most significant digit represents the units, and the remaining digits correspond to the decimal places of the transmittance values. Therefore, to obtain true transmittance values, the calibrated HS image must be divided by 10 4 . Additionally, the software creates a synthetic RGB image, following the methodology explained by Ortega et al. 33 , to facilitate the visualization of the spatial characteristics of the HS image. Prior to any HS image capture, magnification is selected, in this case the 10×. WR and DR reference images are collected. Then, to acquire the HS image, the associated WSI is examined in QuPath 35 to identify an ROI within a specific class, such as IDC (blue), healthy (green) or DCIS (red) tissue. The identified ROI is searched for in the HS microscopic system using the RGB camera and marked down on the histological image using a yellow rectangle (Fig. 2c). The ROI is then captured using the HS microscope, generating the raw HS image, the calibrated HS image, and the synthetic RGB image (Fig. 2d). The RGB image of the ROI is also captured (Fig. 2e) for future analysis. All data corresponding to one of these captures are saved using an identifier with its corresponding metadata, including the patient identifier, classification, and region (e.g., HSI_VNIR_15_IDC_x10_C01; see Data Records section for more details). Data Records The HistologyHSI-BC Recurrence Database has been deposited at The Cancer Imaging Archive (TCIA) repository to make it publicly available, organized into a multilayer folder arrangement 36 . The database is divided into three main components: clinical and demographic data, histological WSI and HS images (see Fig. 3a). The clinical and demographic data are stored at the 00_01_Clinical_Demographic_Data file. This XLSX file documents patients’ demographic status, breast tumor characteristics, treatment received, and their follow-up outcomes (detailed description on Table II). Furthermore, the histological data are structured in 3 folders. Firstly, 01_01_Histological_Images folder contains the WSI for each patient, stored as MRXS files. Each WSI requires a corresponding metadata folder containing DAT and INI files for proper rendering. Moreover, 01_02_Tissue_Annotations folder includes WSI histological annotations that classify tissue types, with boundaries of the regions outlined in blue (IDC), green (healthy), and red (DCIS), as shown in Fig. 2a. The third folder, 01_03_HSI_ROI_Annotations contains the ROI for each HS image, with the boundaries of the region outlined in yellow (Fig. 2a). Both histological WSI and HS image ROI annotation files are provided in GeoJSON format. A summary of the areas of annotations per patient and tissue type is shown in Table III. Lastly, 02_01_HSI_Images folder contains the HS images of the histological slides, stored in ENVI format 37 Each capture includes the raw HS image, WR and DR calibration files, and the calibrated HS image following the procedure described in Equation (1). As the ENVI standard states, the HS image is saved as a flat-binary raster DAT (data) file with an accompanying HDR (header) file containing essential metadata to interpret it. Moreover, within each capture folder a synthetic RGB image and a view of the ROI captured by the RGB camera are stored. The HS image data are stored in folders named according to the regular expression HSI_VNIR_{P}_{TT}_x10_C{CN} , where {P} represents the patient ID, {TT} indicates the tissue type (IDC, healthy, or DCIS), and {CN} is the capture number (Fig. 3b). Table III. Summary of histological area coverage and capture counts by tissue type per image. Patient ID Tissue Annotations Area [mm 2 ] HS image ROI Annotations [Number of captures] Recurrence Label IDC Healthy DCIS IDC Healthy DCIS 15 40.94 2.92 64.24 10 0 0 0 19 94.89 24.00 0.00 10 5 0 1 20 169.27 58.36 0.28 10 5 0 1 25 135.97 139.29 17.62 10 5 5 0 38 210.59 7.71 0.23 10 5 0 1 40 72.19 20.77 0.00 10 5 0 1 43 57.46 22.70 0.00 10 5 0 0 45 71.77 80.06 6.70 10 5 5 0 47 10.40 137.41 0.00 10 5 0 0 51 163.12 2.79 2.34 10 5 0 0 52 96.22 3.50 0.00 10 3 0 0 57 23.26 31.54 0.00 10 5 0 0 62 4.40 66.22 0.00 8 5 0 0 65 104.57 15.22 0.00 10 5 0 0 68 21.41 75.70 0.00 10 5 0 0 70 55.12 45.12 0.00 10 5 0 0 80 4.84 37.06 0.00 7 5 0 0 82 109.32 11.99 0.31 10 5 0 0 84 88.06 13.20 7.50 10 5 0 1 85 119.25 5.21 20.59 10 5 5 0 90 179.57 29.70 0.00 10 5 0 0 100 68.16 3.21 0.06 9 4 0 1 107 8.43 192.86 7.70 10 5 5 0 112 14.84 1.68 0.00 10 0 0 0 124 26.84 0.36 0.00 10 0 0 0 136 190.00 25.98 28.52 9 4 5 0 138 32.58 90.49 11.91 10 5 0 0 139 43.23 72.05 0.00 10 5 0 0 141 168.53 7.96 5.27 10 5 0 1 146 24.16 3.47 0.32 10 0 0 0 151 3.98 3.78 1.10 4 0 0 0 152 82.41 42.77 19.22 10 5 5 1 153 7.06 77.57 0.88 8 5 0 0 154 51.74 0.00 0.00 10 0 0 1 162 67.66 16.19 1.72 10 5 0 1 189 247.15 0.02 0.00 10 0 0 1 197 212.83 57.60 4.90 10 5 5 1 205 321.29 35.69 0.14 10 5 0 1 211 161.00 0.10 0.09 10 0 0 1 213 451.91 7.64 0.00 10 5 0 1 229 281.13 3.09 2.02 10 3 0 1 238 149.90 0.00 0.00 9 0 0 1 255 164.85 35.81 0.50 10 5 0 1 259 92.43 42.47 2.35 10 5 0 1 269 59.97 6.65 0.00 10 5 0 1 270 30.11 111.68 0.00 10 5 0 1 304 155.02 24.54 0.00 10 5 0 1 Technical Validation Clinical and Demographic Data Statistic Analysis A preliminary statistical analysis was conducted to identify differences in the variables between patients with and without recurrence, as shown in Table IV. Statistical tests used for comparisons included the absolute frequency (percentage) for the Chi-square test or Fisher’s exact test, and the median (interquartile range) for the Mann-Whitney U test. As expected, certain classic clinical and pathological variables were found to be associated with the risk of developing metastasis in the present cohort 8 , including age at diagnosis, tumor diameter, and LVI. Table IV. Differences in the clinical and demographic variables in recurrence vs. non-recurrence groups. Attribute Format Recurrence Non-Recurrence p Demographic Data Sex 1: Female 22 (100.0) 25 (100.0) - Race 1: White 22 (100.0) 25 (100.0) - Ethnicity 1: Hispanic 22 (100.0) 25 (100.0) - Menopausal status 0: Premenopause 1: Postmenopause 3 (13.6) 19 (86.4) 3 (12.0) 22 (88.0) 1.000* Diagnosis Dx surgery 0: Mastectomy 1: Lumpectomy 7 (31.8) 15 (68.2) 2 (8.0) 23 (92.0) 0.063* Dx age Years 73.0 [22.0] 57.0 [14.0] 0.017 ‡ Tumor Data Tumor diameter Millimeters 26.5 [13.8] 15.0 [13.0] <0.001 ‡ Tumor histologic grade 1: Grade 1 2: Grade 2 3: Grade 3 1 (4.5) 11 (50.0) 10 (45.5) 4 (16.0) 14 (56.0) 7 (28.0) 0.285* LVI 0: Negative 1: Positive 8 (36.4) 14 (63.6) 18 (72.0) 7 (28.0) 0.031* PNI 0: Negative 1: Positive 15 (68.2) 7 (31.8) 21 (84.0) 4 (16.0) 0.351* Tumor Stage T (tumor) 1: T1 2: T2 3: T3 4: T4 5 (22.7) 14 (63.6) 2 (9.1) 1 (4.5) 17 (68.0) 8 (32.0) 0 (0.0) 0 (0.0) 0.012* N (node) 0: N0 1: N1 2: N2 3: N3 8 (36.4) 7 (31.8) 5 (22.7) 2 (9.1) 25 (100.0) 0 (0.0) 0 (0.0) 0 (0.0) <0.001* M (metastasis) 0: M0 22 (100.0) 25 (100.0) - Tumor Molecular Markers ER 0: Positive ER for < 1% of the TC 1: Positive ER for ≥ 1% of the TC 4 (18.2) 18 (81.8) 5 (20.0) 20 (80.0) 1.000* PR 0: Positive PR for < 1% of the TC 1: Positive PR for ≥ 1% of the TC 6 (27.3) 16 (72.7) 9 (36.0) 16 (64.0) 0.744* HER2 0: Negative 1: Positive 15 (68.2) 7 (31.8) 22 (88.0) 3 (12.0) 0.154* KI67 0: KI67 < 20% 1: KI67 ≥ 20% 4 (18.2) 18 (81.8) 10 (40.0) 15 (60.0) 0.189* Molecular subtype 0: Luminal A 1: Luminal B HER2- 2: Luminal B HER2+ 3: HER2+ 4: Triple negative 4 (18.2) 10 (45.5) 4 (18.2) 3 (13.6) 1 (4.5) 6 (24.0) 13 (52.0) 2 (8.0) 1 (4.0) 3 (12.0) 0.512* Lymph Nodes Status LN status 0: Negative 1: Isolated TC (CC ≤ 0.2 mm) 2: Micrometastasis (0.2 mm 2 mm) 6 (27.3) 2 (9.1) 3 (13.6) 11 (50.0) 22 (88.0) 3 (12.0) 0 (0.0) 0 (0.0) <0.001* LN ITC number Number of lymph nodes with isolated tumor cells 0.0 [0.0] 0.0 [0.0] 0.720 ‡ LN MICRO number Number of lymph nodes with micrometastasis 0.0 [0.0] 0.0 [0.0] 0.027 ‡ LN MACRO number Number of lymph nodes with macrometastasis 0.5 [7.0] 0.0 [0.0] <0.001 ‡ LN number Total number of lymph nodes removed during sentinel lymph node biopsy and/or lymph node dissection 13.0 [15.0] 2.0 [2.0] 0.001 ‡ SLN number Number of sentinel lymph nodes removed during sentinel lymph node biopsy 0.5 [2.0] 2.0 [2.0] <0.001 ‡ SLN status 0: Negative 1: Isolated TC (CC ≤ 0.2 mm) 2: Micrometastasis (0.2 mm 2 mm) 4 (36.4) 2 (18.2) 2 (18.2) 3 (27.3) 22 (88.0) 3 (12.0) 0 (0.0) 0 (0.0) 0.002* Tumor Treatment Tx hormonal 0: Not received 1: Received 5 (22.7) 17 (77.3) 4 (16.0) 21 (84.0) 0.715* Tx CT 0: Not received 1: Received 11 (50.0) 11 (50.0) 14 (56.0) 11 (44.0) 0.906* Tx trastuzumab 0: Not received 1: Received 19 (86.4) 3 (13.6) 23 (92.0) 2 (8.0) 0.654* Tx RT 0: Not received 1: Received 3 (13.6) 19 (86.4) 3 (12.0) 22 (88.0) 1.000* Follow-up Metastasis type 0: No evidence of local or distant metastases 1: Metastasized on nearby tissues or lymph nodes 2: Metastasized in distant organs from primary site 3: Both local and distant metastases are present 0 (0.0) 0 (0.0) 20 (90.9) 2 (9.1) 25 (100.0) 0 (0.0) 0 (0.0) 0 (0.0) <0.001* DFS Months 39.0 [48.0] 150.0 [28.0] <0.001 ‡ Vital status 0: Alive 1: Deceased 2 (9.1) 20 (90.9) 21 (84.0) 4 (16.0) <0.001* Death cause 0: Other causes / Still alive 1: Cancer 5 (22.7) 17 (77.3) 25 (100.0) 0 (0.0) <0.001* OS Months 66.5 [85.0] 150.0 [28.0] <0.001 ‡ Data are expressed as absolute frequency (percentage) for qualitative variables, compared using the Chi-square test or Fisher’s exact text*, and as median [interquartile range] for quantitative variables, analyzed using the Mann-Whitney U test ‡ . CC, cluster cells; CT, chemotherapy; DFS, disease-free survival; Dx, diagnosis; ER, estrogen receptors; HER2, human epidermal growth factor receptor 2; ITC, isolated tumor cells; KI67, proliferation index; LN, lymph node; LVI, lymphovascular invasion; MACRO, macrometastasis; MICRO, micrometastasis; OS, overall survival; PNI, perineural invasion; PR, progesterone receptors; RT, radiotherapy; SLN, sentinel lymph node; TC, tumor cells; Tx, treatment. Analysis of lymph nodes status revealed a significantly higher percentage of micrometastasis and macrometastasis in patients with recurrence compared to those without recurrence. This association remained significant when considering the number of affected lymph nodes. A similar trend was observed in sentinel lymph nodes status, where micrometastasis and macrometastasis were more prevalent in the recurrence group. However, no significant differences were found in lymph nodes containing isolated tumor cells, classified as negative lymph nodes 38 . These findings align with established knowledge that lymph node metastasis is associated with a higher risk of recurrence in BC patients during follow-up 5,39 . Regarding patient follow-up, we confirmed that all patients without recurrence show no evidence of local or distant metastases, whereas patients with recurrence do, with most of them having metastases only in distant organs and a smaller percentage presenting with both local and distant metastases. Among patients with recurrence, 90.9% died, with cancer being the cause of death in 77.3% of cases. In contrast, among the patients without recurrence who died, none died from cancer. As expected, the median disease-free survival (DFS) was significantly shorter in the recurrence group compared to the non-recurrence group, as was overall survival (OS). Histopathology WSIs and Annotation Validation Pathologists qualitatively verified the quality of histopathological slides after the sectioning, processing, and staining phases. They confirmed the absence of artifacts in the ROIs of the WSIs, ensuring that these were not introduced during the tissue preparation or digitization phases. Pathologists did not detect the presence of folds, broken tissues, tears, bubbles, scalpel marks, or bad staining on the ROIs due to the tissue preparation phase. Furthermore, they verified the quality of the digitized histopathological slides making sure there were no issues on the WSIs due to the scanning phase. They confirmed the absence of scanning artifacts like focus issues or white reference problems. The annotations on the WSIs (IDC, healthy, and DCIS) were initially made by one pathologist, with the annotation process then subjected to validation by a second pathologist through a pairwise review. This validation phase played a crucial role in minimizing the inter-observability issue, ensuring consistency and reliability in the annotations. An example of the validated annotations is shown in Fig. 4. HSI Validation A technical validation was performed to ensure the quality of the HistologyHSI-BC Recurrence Database. The HS microscope employed in this study has been thoroughly characterized in previous works 40 , confirming its strong performance for spectral resolution-intensive applications. The system demonstrates a dynamic range of 65.3 ± 0.1 dB in transmittance mode, with a constant dark current of 20 digital numbers, which contributes to a reliable HS image capture. It is capable of capturing 826 spectral bands, providing detailed spectral information and accurately reflecting the spectral properties of the materials under analysis. This is evidenced by a spectral correlation measure of 0.88 ± 0.01 when capturing the WCT-2065 transmittance wavelength calibration standard (Avian Technologies, New London, USA) with a known spectral signature in transmittance mode. The system offers a spatial resolution of 0.739 ± 0.001 µm/pixel, along with a modulation transfer function (MTF) of 370 ± 10 line pairs/mm, ensuring sufficient detail for microscopic imaging. Spatial scanning accuracy is indicated by an eccentricity of 0.04 ± 0.04, and spatial repeatability is shown to have a relative difference of 14 ± 8% across consecutive captures. All values were measured at 10× magnification, the same magnification used for the HS image capturing process in this work. The characterization parameters obtained from the HS microscopic system demonstrate its capability to provide reliable and accurate HS data. The HS images captured from the 47 patients studied underwent a calibration. Afterwards, the database was evaluated to ensure the quality of the captured data. All HS images from each patient and tissue class (IDC, healthy and DCIS) were averaged for visualization purposes. Fig. 5 groups the spectral signatures of patients with and without recurrence after 12 years. Interestingly, in the biopsies, patients without recurrence showed a greater similarity between healthy and DCIS tissues, while these tissue types were more distinctly separated in patients with recurrence. This finding raises the possibility that the closer resemblance of DCIS to healthy tissue could serve as an indicator of non-recurrence. Usage Notes Visualizing histopathology WSIs The authors recommend downloading and installing the QuPath software 35 to visualize and work with the WSIs (MRXS format) and their related annotations (GeoJSON format) (Fig. 2a). There are two ways to open a WSI: drag and drop the MRXS file into QuPath or go on "File/Open" and select and open the MRXS file. There is also a tab on the left side of QuPath's user interface called "Image", in which it is possible to analyze the metadata of the histopathological image, such as width, height, magnification, and resolution. After opening the WSI on QuPath, the two available GeoJSON files containing annotations on the WSI should be imported. One includes the annotations related to the tissue compartments (IDC in red, healthy in green, and DCIS in blue). In contrast, the other defines the ROIs used for capturing the HS images, represented as yellow rectangles. These two files can be opened by dragging and dropping them into QuPath or clicking "File/Import objects from file" and selecting the GeoJSON files. The data from the GeoJSON files is visible by clicking on the tab "Annotations". If the annotation classes are not shown after clicking the "Annotations" tab, click on the button with the three vertical dots on the bottom right of the tab panel, then select "Populate from existing objects/All classes (including sub-classes)" and the class types along with the number of annotations for each will appear. Recommended histopathology WSI processing This section provides guidelines for working with WSIs (MRXS format). Due to their high resolution, efficient processing techniques are necessary to optimize performance and memory usage. Processing high-resolution images can be time-consuming and memory intensive. The highest available resolution of the selected slide image is approximately 85,000 × 202,000 pixels, making it significantly large. To optimize performance, a lower resolution (approximately 670 x 1,600) level should be selected for visualization. It is also important to downscale the annotations to match the selected lower resolution level. Recommended HSI processing After HS data capture, the calibration of HS images is a mandatory step; however, additional processing may be performed depending on the specific application of the data. Given the strong correlation between adjacent spectral bands, spectral dimensionality reduction can be beneficial in reducing intrinsic Gaussian noise and computational costs. This can be accomplished by averaging adjacent spectral bands to create a spectrally reduced HS image. For example, the data could be reduced from the original 826 bands to 275 using a spectral window that includes three neighboring bands. Normalization is also recommended when partial absorbance is less critical, but the specific absorption wavelengths are significant. This normalization can be performed to scale the data between 0 and 1 or to have a mean of 0 and a standard deviation of 1. For HS analysis of the samples, it is advised to remove the sample background by identifying areas with no absorbance, typically represented by the white background. The classification of the data can be based on recurrence status and/or tissue type (IDC, healthy, or DCIS). When using ML or DL, it is crucial to ensure that data from the same patient do not appear simultaneously in the training, testing, or validation sets. How to combine the different databases (Clinical and Demographic, WSIs, and HSI) Integrating the diverse databases within the Histology HSI-BC Recurrence Database can enhance the prediction of distant recurrence in BC by leveraging complementary information from multiple modalities. Histopathological WSIs provide morphological insights assessed by pathologists, HS images capture biochemical variations that may indicate early tumor progression, and clinical and demographic data offer critical patient-specific factors. By integrating these databases, researchers can develop more robust predictive models that go beyond traditional histological or clinical assessments, improving risk stratification and supporting personalized treatment decisions. Researchers can employ various data fusion strategies to achieve this integration. Early fusion involves combining raw or preprocessed features from each modality before model training, allowing the model to learn directly from the integrated data42. Intermediate fusion entails extracting high-level features from each database separately and then merging them into a joint representation, capturing modality-specific patterns prior to integration43. Late fusion consists of training independent models for each modality and subsequently combining their outputs to improve overall prediction accuracy44. Implementing these fusion techniques requires meticulous preprocessing to ensure compatibility and maximize the value of each database. By effectively integrating these multimodal databases, researchers can uncover subtle patterns associated with BC recurrence, advancing precision oncology and personalized patient care. Declarations Code Availability The supplementary files related to this study were created using custom Python scripts. Additionally, several widely used toolboxes and libraries can be accessed via the publicly available repository at https://github.com/HIRIS-Lab/HistologyHSI-BC-Recurrence. A tutorial is provided to show how to manipulate annotations in GeoJSON format by overlaying tissue compartments (IDC, healthy, and DCIS) onto the slide image using their corresponding colors: blue, green, and red. Working with WSIs in Python requires the installation of the OpenSlide library. Additionally, the JSON library is needed for loading annotations, while the PIL library is used to prepare and save slide images. Acknowledgements This work has been supported by the grant Dr. Ferran 2021 (FFPI21/BE01). Moreover, this work was completed while L.Q-Q. was beneficiary of the pre-doctoral grant given by the “Agencia Canaria de Investigación, Innovación y Sociedad de la Información (ACIISI)” of the “Consejería de Economía, Conocimiento y Empleo”, which is part-financed by the European Social Fund (FSE) (POC 2014-2020, Eje 3 Tema Prioritario 74 (85%)). E.S-C. was beneficiary of a grant from the Generalitat de Catalunya (Doctorats Industrials, Pla DI, AGAUR, Grant No. 2022DI057). J.S-N. was supported by the 2024 predoctoral grant from Las Palmas provincial headquarters of the Scientific Foundation of the Spanish Association Against Cancer (PRDLP246561SANT). A.F., a member of the BosomShield project, received funding from Marie Sklodowska-Curie Doctoral Networks Actions (HORIZON-MSCA-2021-DN-01-01) under grant agreement 101073222. C.L-P. was the PI of the SCARLET, a project funded by Proyectos Estrategicos Orientados a la Transición Ecológica y a la Transición Digital, from the 2021 call of the Ministerio de Ciencia e Innovación, with grant number TED2021-130081B-C22 and funding from NextGenerationEU. Moreover, this work is part of the OASIS (Open Ai-driven Stack para plataformas hpec mejoradas en Sistemas Integrado) project (PID2023-148285OB-C43), supported by the Spanish Government and European Union (FEDER Funds). Author contributions L.Q-Q. conceived the study, designed and established the experimental system, carried out a research stay at the Hospital de Tortosa Verge de la Cinta, generated the hyperspectral database, performed image processing and result analysis, and wrote the manuscript. E.S-C. conceived the study, acquired the funding, collected and cleaned the clinical database, generated the histology database, conducted the technical validation experiments, and wrote the manuscript. A.F. generated the histology database and wrote the manuscript. J.S-N. generated the hyperspectral database and wrote the manuscript. S.O. supervised the study and reviewed the manuscript. N.G-B. collected and cleaned the clinical database and reviewed the manuscript. A.F-C. conducted the technical validation experiments and reviewed the manuscript. T.S-A. collected and cleaned the clinical database and reviewed the manuscript. H.F. supervised the study and reviewed the manuscript. L.A-L., D.M-C., and R.B-P. prepared and diagnosed the histological samples, annotated the histopathology images, and reviewed the manuscript. M.L. supervised the study and reviewed the manuscript. G.M.C. and C.L-P. acquired the funding, supervised the study, and reviewed the manuscript. Competing interests The authors declare no conflict of interest. References Bray, F. et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 74 , 229–263 (2024). Chiang, A. C. & Massagué, J. Molecular Basis of Metastasis. N Engl J Med 359 , 2814 (2008). Colzani, E. et al. Time-dependent risk of developing distant metastasis in breast cancer patients according to treatment, age and tumour characteristics. Br J Cancer 110 , 1378–1384 (2014). Siegel, R. L., Giaquinto, A. N. & Jemal, A. Cancer statistics, 2024. CA Cancer J Clin 74 , 12–49 (2024). Riggio, A. I., Varley, K. E. & Welm, A. L. The lingering mysteries of metastatic recurrence in breast cancer. Br J Cancer 124 , 13–26 (2021). Caswell-Jin, J. L. et al. Change in Survival in Metastatic Breast Cancer with Treatment Advances: Meta-Analysis and Systematic Review. JNCI Cancer Spectr 2 , (2018). Iwata, H. Future treatment strategies for metastatic breast cancer: curable or incurable? Breast Cancer 19 , 200–205 (2012). Harbeck, N. et al. Breast cancer. Nature Reviews Disease Primers 2019 5:1 5 , 1–31 (2019). Fasching, P. A. et al. Biomarkers in Patients with Metastatic Breast Cancer and the PRAEGNANT Study Network. Geburtshilfe Frauenheilkd 75 , 41 (2015). Lee, K. K. et al. Prognostic Biomarkers for Breast Cancer Metastasis. Cancer Metastasis (2018) doi:10.5772/INTECHOPEN.80576. Ross, J. S. et al. Immunotherapy predictive biomarkers in metastatic breast cancer (MBC). Journal of Clinical Oncology 37 , 1023–1023 (2019). Prasad, S. N. & Houserkova, D. THE ROLE OF VARIOUS MODALITIES IN BREAST IMAGING. Biomedical Papers 151 , 209–218 (2007). Breast Tumours . World Health Organization Classification of tumours (IACR, Lyon, France, 2019). Jain, E. et al. Whole Slide Imaging Technology and Its Applications: Current and Emerging Perspectives. Int J Surg Pathol 32 , 433–448 (2024). Pantanowitz, L. et al. Twenty Years of Digital Pathology: An Overview of the Road Travelled, What is on the Horizon, and the Emergence of Vendor-Neutral Archives. J Pathol Inform 9 , 40 (2018). Mezei, T., Kolcsár, M., Joó, A. & Gurzu, S. Image Analysis in Histopathology and Cytopathology: From Early Days to Current Perspectives. J Imaging 10 , 252 (2024). Soliman, A., Li, Z. & Parwani, A. V. Artificial intelligence’s impact on breast cancer pathology: a literature review. Diagn Pathol 19 , 38 (2024). Zhao, Y. et al. A review of cancer data fusion methods based on deep learning. Information Fusion 108 , 102361 (2024). Llinas-Bertran, A., Butjosa-Espín, M., Barberi, V. & Seoane, J. A. Multimodal data integration in early-stage breast cancer. The Breast 80 , 103892 (2025). Lipkova, J. et al. Artificial intelligence for multimodal data integration in oncology. Cancer Cell 40 , 1095–1110 (2022). Chang, C.-I. Hyperspectral Imaging . (Springer US, Boston, MA, 2003). doi:10.1007/978-1-4419-9170-6. Fei, B. Hyperspectral imaging in medical applications. in 523–565 (2019). doi:10.1016/B978-0-444-63977-6.00021-3. Karim, S., Qadir, A., Farooq, U., Shakir, M. & Laghari, A. A. Hyperspectral Imaging: A Review and Trends towards Medical Imaging. Curr Med Imaging Rev 19 , 417–427 (2022). Lu, G. & Fei, B. Medical hyperspectral imaging: a review. J Biomed Opt 19 , 010901 (2014). Saccomandi, P. et al. Estimation of anisotropy coefficient of swine pancreas, liver and muscle at 1064 nm based on goniometric technique. J Biophotonics 8 , 422–428 (2015). Wang, Q., Wang, J., Zhou, M., Li, Q. & Wang, Y. Spectral-spatial feature-based neural network method for acute lymphoblastic leukemia cell identification via microscopic hyperspectral imaging technology. Biomed Opt Express 8 , 3017 (2017). Enfield, K. S. S. et al. Hyperspectral cell sociology reveals spatial tumor-immune cell interactions associated with lung cancer recurrence. J Immunother Cancer 7 , 13 (2019). Lingle, W. et al. The Cancer Genome Atlas Breast Invasive Carcinoma Collection (TCGA-BRCA). The Cancer Imaging Archive (2016) doi:10.7937/K9/TCIA.2016.AB2NAZRP. Curtis, C. et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 486 , 346–352 (2012). Fiorin, A., López Pablo, C., Lejeune, M., Hamza Siraj, A. & Della Mea, V. Enhancing AI Research for Breast Cancer: A Comprehensive Review of Tumor-Infiltrating Lymphocyte Datasets. Journal of Imaging Informatics in Medicine 37 , 2996–3008 (2024). Amgad, M. et al. Structured crowdsourcing enables convolutional segmentation of histology images. Bioinformatics 35 , 3461–3467 (2019). Zhang, Y., Wang, Y., Zhang, B. & Li, Q. A hyperspectral dataset of precancerous lesions in gastric cancer and benchmarks for pathological diagnosis. J Biophotonics 15 , (2022). Ortega, S. et al. Publisher Correction: Histological Hyperspectral Glioblastoma Dataset (HistologyHSI-GB). Sci Data 11 , 829 (2024). WHO Classification of Tumours of the Breast . (IARC, Lyon, France, 2012). Bankhead, P. et al. QuPath: Open source software for digital pathology image analysis. Sci Rep 7 , 16878 (2017). Quintana-Quintana, L. et al. HistologyHSI-BC Recurrence Database. Preprint at (2025). ENVI Header Files. https://www.nv5geospatialsoftware.com/docs/ENVIHeaderFiles.html#The. Naidoo, K. & Pinder, S. E. Micro- and macro-metastasis in the axillary lymph node: A review. The Surgeon 15 , 76–82 (2017). Jatoi, I., Hilsenbeck, S. G., Clark, G. M. & Osborne, C. K. Significance of Axillary Lymph Node Metastasis in Primary Breast Cancer. Journal of Clinical Oncology 17 , 2334–2334 (1999). Quintana-Quintana, L. et al. Roadmap for the Characterization and Validation of Hyperspectral Microscopic Systems [submitted]. IEEE Trans Instrum Meas 1–16 (2025) doi:10.1109/TIM.2024.3379090. Wang, Z. et al. Deep learning-based multi-modal data integration enhancing breast cancer disease-free survival prediction. Precis Clin Med 7 , (2024). Nakach, F.-Z., Idri, A. & Goceri, E. A comprehensive investigation of multimodal deep learning fusion strategies for breast cancer classification. Artif Intell Rev 57 , 327 (2024). Llinas-Bertran, A., Butjosa-Espín, M., Barberi, V. & Seoane, J. A. Multimodal data integration in early-stage breast cancer. The Breast 80 , 103892 (2025). Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7242335","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":492575345,"identity":"b22445b5-d3ba-41fe-9b80-22aefb69b24d","order_by":0,"name":"Laura Quintana-Quintana","email":"","orcid":"https://orcid.org/0000-0003-1154-6490","institution":"Institute for Applied Microelectronics, University of Las Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain.","correspondingAuthor":false,"prefix":"","firstName":"Laura","middleName":"","lastName":"Quintana-Quintana","suffix":""},{"id":492575346,"identity":"8af41fe8-1ee2-4229-9f75-269beabedb8e","order_by":1,"name":"Esther Sauras-Colón","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0003-1649-938X","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":true,"prefix":"","firstName":"Esther","middleName":"","lastName":"Sauras-Colón","suffix":""},{"id":492575347,"identity":"8005b906-97b6-463f-bdb7-5c42bc6ffd6e","order_by":2,"name":"Alessio Fiorin","email":"","orcid":"https://orcid.org/0009-0002-0315-1085","institution":"Department of Computer Engineering and Mathematics, Universitat Rovira i Virgili, Tarragona, Spain, Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Alessio","middleName":"","lastName":"Fiorin","suffix":""},{"id":492575348,"identity":"36fc899a-cd67-4d2a-a664-bf1be104b35d","order_by":3,"name":"Javier Santana-Nunez","email":"","orcid":"https://orcid.org/0009-0000-5029-5239","institution":"Research Unit, Hospital Universitario de Gran Canaria Dr. Negrín, 35019 Las Palmas de Gran Canaria, Spain; Fundación Canaria Instituto de Investigación Sanitaria de Canarias (FIISC), Las Palmas de Gran Canaria, Spain and Institute for Applied Microelectronics, University of Las Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain.","correspondingAuthor":false,"prefix":"","firstName":"Javier","middleName":"","lastName":"Santana-Nunez","suffix":""},{"id":492575349,"identity":"b4aa77b0-4bad-46c5-ac01-66990335188f","order_by":4,"name":"Samuel Ortega","email":"","orcid":"https://orcid.org/0000-0002-7519-954X","institution":"Norwegian Institute of Food, Fisheries and Aquaculture Research (Nofima), Tromsø, Norway and Institute for Applied Microelectronics, University of Las Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain.","correspondingAuthor":false,"prefix":"","firstName":"Samuel","middleName":"","lastName":"Ortega","suffix":""},{"id":492575350,"identity":"84252ee4-a799-41e6-9c08-9915ba40a291","order_by":5,"name":"Noèlia Gallardo-Borràs","email":"","orcid":"https://orcid.org/0000-0001-8537-4586","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Noèlia","middleName":"","lastName":"Gallardo-Borràs","suffix":""},{"id":492575351,"identity":"dd8bef4f-71a7-499e-8ed1-940042664c34","order_by":6,"name":"Alba Fischer-Carles","email":"","orcid":"https://orcid.org/0009-0005-7733-015X","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain; Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain and BCN MedTech, Department of Engineering, Universitat Pompeu Fabra, Barcelona, Spain","correspondingAuthor":false,"prefix":"","firstName":"Alba","middleName":"","lastName":"Fischer-Carles","suffix":""},{"id":492575352,"identity":"e96793fb-fe30-4c6e-92bf-4e7e1f3f4721","order_by":7,"name":"Tábata Sánchez-Alcántara","email":"","orcid":"https://orcid.org/0009-0002-8797-5347","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Tábata","middleName":"","lastName":"Sánchez-Alcántara","suffix":""},{"id":492575353,"identity":"ac3ff589-64f7-4a94-ad13-4fa528c83685","order_by":8,"name":"Himar Fabelo","email":"","orcid":"https://orcid.org/0000-0002-9794-490X","institution":"Research Unit, Hospital Universitario de Gran Canaria Dr. Negrín, 35019 Las Palmas de Gran Canaria, Spain; Fundación Canaria Instituto de Investigación Sanitaria de Canarias (FIISC), Las Palmas de Gran Canaria, Spain and Institute for Applied Microelectronics, University of Las Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain.","correspondingAuthor":false,"prefix":"","firstName":"Himar","middleName":"","lastName":"Fabelo","suffix":""},{"id":492575354,"identity":"eb4433f3-1d3e-4b36-af86-74ac2c11c05a","order_by":9,"name":"Laia Adalid-Llansa","email":"","orcid":"https://orcid.org/0000-0001-7543-9921","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Laia","middleName":"","lastName":"Adalid-Llansa","suffix":""},{"id":492575355,"identity":"8d87aade-8035-4d56-bd06-42ec73dde4c7","order_by":10,"name":"Daniel Mata-Cano","email":"","orcid":"https://orcid.org/0000-0001-7543-4856","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Mata-Cano","suffix":""},{"id":492575356,"identity":"5762efd3-a9cf-4b7b-b312-641f3cb88332","order_by":11,"name":"Ramon Bosch-Príncep","email":"","orcid":"https://orcid.org/0000-0003-4104-5515","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Ramon","middleName":"","lastName":"Bosch-Príncep","suffix":""},{"id":492575357,"identity":"09892fc1-b207-46ce-8b2c-b4e098447758","order_by":12,"name":"Marylène Lejeune","email":"","orcid":"https://orcid.org/0000-0001-8441-9404","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Marylène","middleName":"","lastName":"Lejeune","suffix":""},{"id":492575358,"identity":"787d471c-410f-4337-855d-19885e10cd97","order_by":13,"name":"Gustavo M. Callico","email":"","orcid":"https://orcid.org/0000-0002-3784-5504","institution":"Institute for Applied Microelectronics, University of Las Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain.","correspondingAuthor":false,"prefix":"","firstName":"Gustavo","middleName":"M.","lastName":"Callico","suffix":""},{"id":492575359,"identity":"69d68a6a-0ed1-4267-9904-f7529f43fe11","order_by":14,"name":"Carlos López-Pablo","email":"","orcid":"https://orcid.org/0000-0003-1248-3065","institution":"Department of Pathology, Hospital de Tortosa Verge de la Cinta, Institut Català de la Salut, Tortosa, Spain and Oncological Pathology and Bioinformatics Research Group, Institut d'Investigació Sanitària Pere Virgili, Tortosa, Spain","correspondingAuthor":false,"prefix":"","firstName":"Carlos","middleName":"","lastName":"López-Pablo","suffix":""}],"badges":[],"createdAt":"2025-07-29 10:44:54","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":true,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-7242335/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7242335/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":87903436,"identity":"3906ea09-ecaf-4fec-8dd2-d5a78aa654cb","added_by":"auto","created_at":"2025-07-30 08:33:19","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":135024,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic overview of the study design or workflow. The patient cohort comprises 47 BC patients, of which 22 (47%) experienced distant recurrence over a 12-year follow-up period and 25 (53%) did not. All these patients had a breast tumor biopsy, which was collected, prepared, and stained with H\u0026amp;E. The H\u0026amp;E-stained slides were digitized to obtain WSIs and annotated to differentiate three tissue compartments: IDC, healthy tissue and DCIS. Within each annotated area, ROIs were selected, from which HS images were acquired to generate HS cubes and extract spectral signatures. Together with the clinical and demographic data, all this information constitutes the HistologyHSI - BC Recurrence Database.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7242335/v1/9452c29068a12727c37d5824.png"},{"id":87902763,"identity":"fc82fb87-ee33-4b0e-959d-dde5e738527a","added_by":"auto","created_at":"2025-07-30 08:25:19","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":457196,"visible":true,"origin":"","legend":"\u003cp\u003eElements and outputs to capture an HS image. a) Annotated WSI (IDC outlined in blue, healthy tissue in green, and DCIS in red) captured with the WSI scanner. b) HS microscopic system used to acquire an HS image and its corresponding high spatial resolution RGB image of a selected ROI. c) Zoom-in of one of the selected ROIs outlined by yellow rectangles in a). d) Synthetic RGB image generated from the captured HS image. e) High spatial resolution RGB image captured using the 20 MP RGB camera of the HS microscopic system.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7242335/v1/8aef66a8d1c182ec16940083.png"},{"id":87902761,"identity":"ad957dfd-54fe-4ebf-882d-cc25b8e6c112","added_by":"auto","created_at":"2025-07-30 08:25:19","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":47041,"visible":true,"origin":"","legend":"\u003cp\u003eGraphical representation of a) the HistologyHSI-BC Recurrence Database structure and b) the contents of each HS image capture.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7242335/v1/a4d9727b9d417875316a028a.png"},{"id":87902767,"identity":"0af4741b-5c6d-43e2-81cb-bdac487d131a","added_by":"auto","created_at":"2025-07-30 08:25:19","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":189552,"visible":true,"origin":"","legend":"\u003cp\u003eExamples of (a) IDC, (b) healthy, and (c) DCIS tissue types on a WSI at 2x magnification.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7242335/v1/97b952442bd4e1de4ddc6f80.png"},{"id":87902766,"identity":"fc06c650-90bc-4371-9719-5444358743e7","added_by":"auto","created_at":"2025-07-30 08:25:19","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":58084,"visible":true,"origin":"","legend":"\u003cp\u003eMean and standard deviation HS spectral signatures for different tissues (IDC, healthy and DCIS) for patients a) without and b) with recurrence.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7242335/v1/d118a3eb6bc1d49dd2faedbb.png"},{"id":87904766,"identity":"fe6a39c7-09b3-45f8-8bd5-955fb2f58a39","added_by":"auto","created_at":"2025-07-30 08:41:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2690697,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7242335/v1/c1f89774-60a3-4e5a-9b47-0a95012fde56.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eHistological Hyperspectral Breast Cancer Recurrence Database (HistologyHSI-BC Recurrence)\u003c/p\u003e","fulltext":[{"header":"Background \u0026 Summary","content":"\u003cp\u003eIn 2022, breast cancer (BC) was the most common type of cancer in women, with an incidence of 23.8%, and the leading cause of cancer-related death among women, accounting for 15.4% of all cancer-related deaths\u003csup\u003e1\u003c/sup\u003e. Cancer cells can spread from the primary tumor to other parts of the body, which is known as metastasis and is the main cause of death in most cancers\u003csup\u003e\u003cspan class=\"MsoCommentReference\"\u003e2,3\u003c/span\u003e\u003c/sup\u003e. Metastasis occurs in nearly 1 out of 3 patients diagnosed with BC and can appear in the axillary lymph nodes (regional metastasis) or in other organs (distant metastasis)\u003csup\u003e4\u003c/sup\u003e. Overall, the 5-year survival rate after diagnosis of BC is 91%. However, this rate is higher in patients with tumors located exclusively in the breast (99%) than in patients with regional metastasis (86%) or with distant metastasis (31%), where survival decreases dramatically\u003csup\u003e4\u003c/sup\u003e. Women with distant metastases may have either \u003cem\u003ede novo\u003c/em\u003e distant spread, where distant metastases are already present at the time of diagnosis, or develop distant metastases after an initial diagnosis and treatment, which is known as \u003cem\u003erecurrence\u003c/em\u003e\u003csup\u003e5\u003c/sup\u003e. While de novo cases account for approximately 25% of metastatic BC diagnoses, the majority result from recurrence\u003csup\u003e6\u003c/sup\u003e. Whether de novo or recurrent, distant metastases remain incurable\u003csup\u003e7,8\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eCertain classic prognostic factors are associated with the risk of developing distant metastasis, such as \u003cem\u003eage\u003c/em\u003e, \u003cem\u003etumor diameter\u003c/em\u003e, \u003cem\u003estage\u003c/em\u003e, \u003cem\u003etumor grade\u003c/em\u003e, \u003cem\u003etumor type\u003c/em\u003e or \u003cem\u003elymphovascular invasion\u003c/em\u003e (LVI)\u003csup\u003e8\u003c/sup\u003e. Additionally, studies have identified other biomarkers with prognostic value in the disease that may be associated with metastasis, including genetic alterations, circulating tumor cells and circulating tumor DNA, biomarkers of response to immunotherapy and gene expression platforms to predict the risk of recurrence\u003csup\u003e9\u0026ndash;11\u003c/sup\u003e. However, to date there is no consensus for the implementation of most of these biomarkers in routine clinical practice. Therefore, there continues to be a growing interest in identifying specific prognostic biomarkers that allow determining the probability of developing metastasis.\u003c/p\u003e\n\u003cp\u003eCancer detection relies heavily on imaging methods like X-ray, ultrasound, and magnetic resonance imaging\u003csup\u003e12\u003c/sup\u003e. However, treatment decisions require a conclusive histopathological diagnosis, which is obtained from a tissue biopsy. BC can be broadly categorized into in situ carcinoma and invasive carcinoma. Among these, ductal carcinoma in situ (DCIS) represents the most prevalent subtype of in situ carcinoma, while invasive ductal carcinoma (IDC) is the most common subtype of invasive carcinoma. Nevertheless, given the heterogeneity of BC, the accurate identification of these subtypes among other histological subtypes requires extensive expertise and a deep understanding of breast pathology\u003csup\u003e13\u003c/sup\u003e.\u0026nbsp;The rise of digital pathology, which leverages whole-slide images (WSIs), has revolutionized research and diagnosis in pathology, particularly in cancer, by enabling more efficient data sharing across institutions and promoting remote collaborations. WSIs are high-resolution digital images of traditional \u0026nbsp;glass pathology slides, which can be viewed, analyzed, and shared on a computer screen\u003csup\u003e14,15\u003c/sup\u003e. The use of WSIs also paves the way for computational pathology, which started from the use of traditional image analysis methods to advanced machine learning (ML) and deep learning (DL) algorithms\u003csup\u003e16,17\u003c/sup\u003e. Remarkably, these novel approaches offer the potential to integrate multiple data modalities, extending beyond histopathology image analysis. This includes linking histopathological images with clinical factors, such as prognosis and genetic mutations, thereby enhancing BC diagnostics\u003csup\u003e18\u0026ndash;20\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eBeyond conventional methods, other imaging modalities show promising potential for improving the diagnosis and prognosis of BC patients. Among these, hyperspectral (HS) imaging (HSI), combines traditional imaging with spectroscopy to capture both spatial and spectral information. Each material interacts uniquely with emitted radiation, reflecting and absorbing it in a way that creates a distinct radiance vector, often named \u003cem\u003espectral signature\u003c/em\u003e. HSI sensors can capture these spectral signatures, acquiring significantly more data than standard RGB (Red, Green, Blue) cameras and extending imaging capabilities beyond human vision (e.g., near-infrared (NIR) HS sensors can capture wavelengths ranging from 900 to 1,700 nm)\u003csup\u003e21\u003c/sup\u003e. In recent years, the use of HSI in medicine has begun to achieve promising results regarding cancer detection by utilizing cutting-edge ML algorithms to process the high amount of HS data\u003csup\u003e22\u0026ndash;24\u003c/sup\u003e.\u0026nbsp;In the existing literature for medical histological applications, HSI has been used to identify pancreatic neoplasms with different prognoses\u003csup\u003e25\u003c/sup\u003e, quantify Ki67 as a prognostic factor in lymphomas\u003csup\u003e26\u003c/sup\u003e and study the interactions between tumor cells and immune cells of the tumor microenvironment in response to immunotherapy in lung cancer\u003csup\u003e27\u003c/sup\u003e, obtaining promising results.\u003c/p\u003e\n\u003cp\u003eResearch in this area is still in its early stages, and the number of published studies remains limited. Regarding WSI and clinical and demographic databases, The Cancer Genome Atlas (TCGA) is one of the main publicly available sources for hematoxylin and eosin (H\u0026amp;E)-stained WSIs and associated clinical and demographic data\u003csup\u003e28\u003c/sup\u003e. The main challenge of this database is the lack of annotations, which makes the subsequent analysis of these WSIs difficult. The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) is another database that provides clinical, demographic, and molecular data of over 2,000 BC cases\u003csup\u003e29\u003c/sup\u003e. However, it primarily focuses on genomic and transcriptomic data and does not include WSIs. Several publicly available databases contain H\u0026amp;E-stained WSIs\u003csup\u003e30\u003c/sup\u003e, but most include annotations focused on specific cell types in BC, such as tumor-infiltrating lymphocytes and inflammatory cells. We have identified one database - Breast Cancer Semantic Segmentation (BCSS) that provides specific annotations of tissue compartments, distinguishing between tumor and healthy tissue\u003csup\u003e31\u003c/sup\u003e. Regarding medical HS databases only two major sets were identified: in 2022, Zhang \u003cem\u003eet al.\u003c/em\u003e\u003csup\u003e32\u003c/sup\u003e introduced a large-scale database for HS microscopic images of precancerous lesions in gastric cancer, and in 2024, Ortega \u003cem\u003eet al.\u0026nbsp;\u003c/em\u003ereleased the HistologyHSI-GB dataset\u003csup\u003e33\u003c/sup\u003e, focused on HS glioblastoma histology. None of those databases included clinical or demographic data, this being one of the main challenges in this field. There is a limited availability of comprehensive, high-quality databases, which hinders the broader application of clinical and demographic data, WSIs and HSI in clinical practice and research. The lack of such databases makes it difficult to fully explore the potential of HSI together with conventional practices for diagnosing and predicting disease outcomes, such as recurrence in BC.\u003c/p\u003e\n\u003cp\u003eThis paper presents a publicly accessible database designed to investigate specific prognostic biomarkers for predicting the likelihood of BC recurrence due to distant metastasis. The HistologyHSI-BC Recurrence Database includes clinical and demographic data from BC patients, along with WSIs and HS images obtained from their primary tumor samples. This database is intended to evaluate the ability to predict recurrence due to distant metastasis over a 12-year follow-up period. Biopsies from 47 patients diagnosed with BC were extracted, sliced and stained with H\u0026amp;E, 47 WSIs and 677 microscopic HS images were taken, and their clinical and demographic data were collected. Among these patients, 22 experienced distant recurrence. A schematic overview of the study workflow is presented in Fig. 1.\u003c/p\u003e\n\u003cp\u003eThe HistologyHSI-BC Recurrence Database will benefit researchers by providing a comprehensive, multimodal database that integrates WSIs, HS images, and clinical and demographic data from BC patients. This resource enables the development and refinement of predictive models for BC recurrence due to distant metastasis, starting to fill a significant gap in available databases. Researchers can leverage this data to explore innovative ML approaches, enhance diagnostic accuracy, and identify novel biomarkers for BC recurrence. Additionally, the database promotes reproducibility, facilitates collaboration across institutions, and accelerates research in computational pathology, potentially improving personalized treatment strategies and benefiting broader cancer research efforts.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003ePatients Selection, Eligibility Criteria and Ethics Approval\u003c/h2\u003e\n\u003cp\u003eThis is a retrospective case-control study carried out on 47 BC patients diagnosed with IDC, now called invasive breast carcinoma of no special type\u003csup\u003e34\u003c/sup\u003e, between 2006 and 2015. Cases include 22 patients who experienced recurrence due to distant metastasis during the 12 years following diagnosis. The remaining 25 patients who did not experience recurrence during the 12 years of follow-up are included as control group.\u003c/p\u003e\n\u003cp\u003eTable I. Eligibility criteria for patient inclusion.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInclusion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eExclusion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003eA diagnosis of IDC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003eReceipt of neoadjuvant treatment, as it is known to modify the tumor microenvironment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003eRepresentativeness of IDC tissue in surgical biopsy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003eRecurrence occurring in the breast rather than in distant organs\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003eA clinical history with complete clinical and pathological data\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003ePresence of distant metastases at the time of diagnosis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003ePatient\u0026rsquo;s agreement to be included in the study. The exclusion criteria included\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50%;\"\u003e\n \u003cp\u003eFailure to meet any of the inclusion criteria\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eThe study was approved by the Drug Research Ethics Committee of the Institut d\u0026rsquo;Investigaci\u0026oacute; Sanit\u0026agrave;ria Pere Virgili (IISPV), Tarragona, Spain, under reference number 134/2022. The samples used in this study were obtained from Biobank IISPV-Node Tortosa, Tarragona, Spain, following the principles of ethical conduct and data protection. All participants whose samples were stored in the biobank have previously signed an informed consent form, explicitly authorizing the collection, storage, and future use of their biological materials and associated data for research purposes. The processing, communication and transfer of personal data of all participants comply with the provisions of Organic Law 3/2018, of December 5, on the Protection of Personal Data and Guarantee of Digital Rights and with Regulation (EU) 2016/679 of the European Parliament and of the Council, of April 27, 2016, on the protection of natural persons with regard to the processing of personal data and the free circulation of these data, and repealing Directive 95/46/EC (General Data Protection Regulation). \u0026nbsp;The data generated and collected during this study are anonymized to ensure the security of the information, safeguarding the confidentiality and privacy of the patients.\u003c/p\u003e\n\u003ch2\u003eData Collection\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eClinical and Demographic Data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data collection process involved extracting information from clinical records, including demographic and clinical data, which were following Table II.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;II. Description of the study variables.\u003c/p\u003e\n\u003cdiv align=\"center\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"555\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAttribute\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDefinition\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFormat\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDemographic Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatients\u0026apos; gender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e1: Female\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRace\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatients\u0026rsquo; race\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e1: White\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthnicity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatients\u0026rsquo; ethnicity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e1: Hispanic\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMenopausal\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eMenopausal status of the patient\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Premenopause\u003c/p\u003e\n \u003cp\u003e1: Postmenopause\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiagnosis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDx\u0026nbsp;surgery\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eType of surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Mastectomy\u003c/p\u003e\n \u003cp\u003e1: Lumpectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDx\u0026nbsp;age\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eDifference between the date of diagnosis and the date of birth\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eYears\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor\u0026nbsp;diameter\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eMaximum diameter of the irregular shaped tumor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eMillimeters\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor\u0026nbsp;histologic\u0026nbsp;grade\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eDegree of differentiation of tumor cells, reflecting how different they are from normal breast cells\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e1: Grade 1\u003c/p\u003e\n \u003cp\u003e2: Grade 2\u003c/p\u003e\n \u003cp\u003e3: Grade 3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eL\u003c/strong\u003e\u003cstrong\u003eymphovascular Invasion (L\u003c/strong\u003e\u003cstrong\u003eVI\u003c/strong\u003e\u003cstrong\u003e)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePresence of tumor cells within lymphatic or blood vessels\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Positive\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eP\u003c/strong\u003e\u003cstrong\u003eerineural Invasion (P\u003c/strong\u003e\u003cstrong\u003eNI\u003c/strong\u003e\u003cstrong\u003e)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eAbility of cancer cells to proliferate around peripheral nerves and, eventually, invade them\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Positive\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Stage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eT (tumor)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eTumor size assessed by pathological evaluation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e1: T1\u003c/p\u003e\n \u003cp\u003e2: T2\u003c/p\u003e\n \u003cp\u003e3: T3\u003c/p\u003e\n \u003cp\u003e4: T4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN (node)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eThe cancer has spread to the lymph nodes assessed by pathological evaluation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: N0\u003c/p\u003e\n \u003cp\u003e1: N1\u003c/p\u003e\n \u003cp\u003e2: N2\u003c/p\u003e\n \u003cp\u003e3: N3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eM (metastasis)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eStatus of breast cancer spreading to a different part of the body\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: M0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Molecular Markers\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(determined by IHC)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eER\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eStatus of estrogen receptors (ER)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Positive ER for \u0026lt; 1% of the TC\u003c/p\u003e\n \u003cp\u003e1: Positive ER for \u0026ge; 1% of the TC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eStatus of progesterone receptors (PR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Positive PR for \u0026lt; 1% of the TC\u003c/p\u003e\n \u003cp\u003e1: Positive PR for \u0026ge; 1% of the TC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHER2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eStatus of human epidermal growth factor receptor 2 (HER2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Negative\u0026nbsp;(IHC = 0/1+, or IHC = 2+ and FISH = negative)\u003c/p\u003e\n \u003cp\u003e1: Positive\u0026nbsp;(IHC = 3+, or IHC = 2+ and FISH = positive)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eKI67\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eIndex quantifying KI67 expression to measure how fast cancer cells are dividing in a tumor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: KI67 index \u0026lt; 20%\u003c/p\u003e\n \u003cp\u003e1: KI67 index \u0026ge; 20%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMolecular\u0026nbsp;subtype\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eClassification according to IHC status of ER, PR, HER2 and Ki67.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Luminal A\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(ER+/PR+, HER2-, KI67-)\u003c/p\u003e\n \u003cp\u003e1: Luminal B HER2-\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(ER+/PR+, HER2-, KI67+)\u003c/p\u003e\n \u003cp\u003e2: Luminal B HER2+\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(ER+/PR+, HER2+, KI67-/+)\u003c/p\u003e\n \u003cp\u003e3: HER2+\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(ER-, PR-, HER2+, KI67-/+)\u003c/p\u003e\n \u003cp\u003e4: Triple negative\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(ER-, PR-, HER2-, KI67-/+)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"7\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLymph Nodes Status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eStatus of the spreading of tumor cells to the lymph nodes, including sentinel and non-sentinel lymph nodes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Isolated TC (CC \u0026le; 0.2 mm)\u003c/p\u003e\n \u003cp\u003e2: Micrometastasis (0.2 mm \u0026lt; CC \u0026le; 2 mm)\u003c/p\u003e\n \u003cp\u003e3: Macrometastasis (CC \u0026gt; 2 mm)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;ITC\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eLymph nodes with isolated tumor cells\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eNatural number\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;MICRO\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eLymph nodes with micrometastasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eNatural number\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;MACRO\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eLymph nodes with macrometastasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eNatural number\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eLymph nodes removed during sentinel lymph node biopsy and/or lymph node dissection\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eNatural number\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSLN\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eLymph nodes removed during sentinel lymph node biopsy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eNatural number\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSLN\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePresence (or absence) of tumor cells in the sentinel lymph node(s), which are the first lymph node(s) where cancer cells might spread from a tumor.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Isolated TC (CC \u0026le; 0.2 mm)\u003c/p\u003e\n \u003cp\u003e2: Micrometastasis (0.2 mm \u0026lt; CC \u0026le; 2 mm)\u003c/p\u003e\n \u003cp\u003e3: Macrometastasis (CC \u0026gt; 2 mm)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Treatmen\u003c/strong\u003e\u003cstrong\u003et\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;hormonal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatient received (or not) hormonal treatment to stop estrogen and progesterone from attaching to their receptors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;CT\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatient received (or not) adjuvant chemotherapy (CT) after the surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;trastuzumab\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatient received (or not) trastuzumab, a targeted therapy used to treat HER2+ breast cancer (Molecular_subtype = 3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;RT\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003ePatient received (or not) radiotherapy (RT)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" style=\"width: 36px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFollow-up\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMetastasis\u0026nbsp;type\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eStatus of cancer spreading from the primary tumor to other organs during the follow-up period\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: No evidence of local or distant metastases\u003c/p\u003e\n \u003cp\u003e1: Metastasized on nearby tissues or lymph nodes\u003c/p\u003e\n \u003cp\u003e2: Metastasized in distant organs from primary site\u003c/p\u003e\n \u003cp\u003e3: Both local and distant metastases are present\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDFS\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eThe time a patient survives without any signs or symptoms of cancer after finishing primary treatment. It is calculated as the difference between the date of relapse and the date of diagnosis. If the patient did not relapse, the date of last follow-up is used instead.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eMonths\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVital\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Alive\u003c/p\u003e\n \u003cp\u003e1: Deceased\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDeath\u0026nbsp;cause\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003e0: Other causes / Still alive\u003c/p\u003e\n \u003cp\u003e1: Cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 216px;\"\u003e\n \u003cp\u003eThe time from the date of cancer diagnosis that patients remain alive. It is calculated as the difference between the date of death and the date of diagnosis. If the patient did not die, the date of last follow-up is used instead.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 200px;\"\u003e\n \u003cp\u003eMonths\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eCC, cluster cells; CT, chemotherapy; DFS, disease-free survival; Dx, diagnosis; ER, estrogen receptors; FISH, fluorescence in situ hybridization; HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry; ITC, isolated tumor cells; KI67, proliferation index; LN, lymph node; LVI, lymphovascular invasion; MACRO, macrometastasis; MICRO, micrometastasis; OS, overall survival; PNI, perineural invasion; PR, progesterone receptors; RT, radiotherapy; SLN, sentinel lymph node; TC, tumor cells; Tx, treatment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHistopathology WSIs\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParaffin blocks of primary tumor biopsies with sufficient representative IDC tissue were obtained from the Biobank IISPV-Node Tortosa, Tarragona, Spain. The samples were processed in the Pathology Department, where 2 \u0026micro;m-thick sections were prepared from each paraffin block and stained with H\u0026amp;E according to the instructions of the manufacturer. The slides were sealed with coverslips using dibutylphthalate polystyrene xylene (DPX) mounting medium for subsequent digitization and HS microscopic image acquisition.\u003c/p\u003e\n\u003cp\u003eThe H\u0026amp;E-stained slides were digitized with the Pannoramic 250 Flash III WSI scanner (3DHISTECH Ltd., Budapest, Hungary) at 20\u0026times; magnification (0.2433 \u0026micro;m/pixel) using MRXS image format. WSIs were visualized using QuPath\u003csup\u003e35\u003c/sup\u003e for technical validation and annotation, and anonymized using the SlideMaster software (3DHISTECH Ltd., Budapest, Hungary). The annotation process of each WSI was manually performed by pathologists using diverse colors to distinguish between IDC, healthy tissue, and DCIS. The annotations were made with the following color scheme: IDC was outlined in blue, healthy tissue in green, and DCIS in red\u0026nbsp;(Fig. 2a).\u003c/p\u003e\n\u003cp\u003eMoreover, within each of these three tissue compartment types, different regions of interest (ROIs), surrounded by yellow line, were identified and annotated to subsequently acquire the HS image using the HS microscopic system (Fig. 2a). These ROIs were selected to ensure the inclusion of representative areas of each class (IDC, healthy and DCIS), capturing relevant spectral variability for further analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHS Images\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe HS images were captured using a HS microscopic system (Fig. 2b). The system features the Hyperspec\u0026reg; VNIR (Visible and Near Infrared) A-Series camera (HeadWall Photonics, Fitchburg, MA, USA), a pushbroom HS camera that captures data by scanning the sample spatially. The camera is equipped with a charge-coupled cevice (CCD) sensor that covers a spectral range of 400-1,000 nm, capturing 826 spectral bands across 1,004 spatial pixels per line. It offers high spectral resolution with a slit image full width at half maximum (FWHM) of 2.5 nm and a pixel size of 7.4 \u0026mu;m. Data are acquired with a 12-bit ADC (Analog-to-digital Converter), and each HS line has a size of 1,004 \u0026times; 826 pixels and requires 1.6 MB per line on disk for storage. The microscope used is the OLYMPUS BX-53 (Olympus, Tokyo, Japan), with LMPLN-IR (5\u0026times;, 10\u0026times;) and LCPLN-IR (20\u0026times;, 50\u0026times;) objective lenses optimized for infrared imaging. The system uses a 100 W TH4 halogen lamp (Olympus, Tokyo, Japan) as the light source, covering a wavelength range from 400 to 1,800 nm and supporting both transmittance and reflectance light paths. To acquire full HS images, the pushbroom camera requires spatial scanning, which is facilitated by a SCAN 130 \u0026times; 85 scanning stage (M\u0026auml;rzh\u0026auml;user, Wetzlar, Germany). The stage ensures high precision (\u0026plusmn;3 \u0026mu;m accuracy) as it moves the sample, keeping it aligned with the objective and light path. Furthermore, an RGB camera, the acA5472-17uc (Basler AG, Ahrensburg, Germany), provides real-time visualization of the sample to navigate it without the need of using the microscope binoculars. It is a 20 MP compact camera with a Sony IMX183 CMOS sensor (Tokyo, Japan), 5,496 \u0026times; 3,672 resolution, and 17 fps. It features USB 3.0, a C-mount, and supports hardware/software triggers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCalibration of the HS images is necessary to ensure the data accurately represents the sample\u0026apos;s spectral signatures. The HS microscope captures spectral signatures for each pixel, but factors like the sensor\u0026apos;s response, light transmission, and the light source can affect accuracy. The calibration process involves normalizing the pixel values of the HS image by adjusting them based on a white reference (WR) and a dark reference (DR). WR is obtained by focusing on an empty area of the slide at the same working distance. This ensures no sample material is present, allowing the frame to record the maximum signal the sensor can measure for each pixel and wavelength under the given conditions (e.g., exposure time, light intensity, and slide properties). Conversely, the DR is captured by completely blocking light transmission to the HS camera. This frame captures the minimum signal levels detectable by the sensor for each pixel and band, as well as dark current information from the CCD. Ideally, DR values approach zero; however, higher values may occur due to intrinsic sensor noise. To enhance the reliability of the calibration process, 100 frames are collected for both the WR and DR, ensuring that averaging reduces potential errors. Finally, the calibration of the HS image is achieved using Equation (1), which relates the calibrated HS image (ri ) to the raw HS image (Rawi ).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003cp\u003eIn-house software was developed to serve multiple functions in the HSI acquisition process. It displays the RGB image to facilitate sample navigation under the microscope and ensures synchronization between the HS camera and the scanning platform by aligning their frame rate and platform movement. After capturing the HS image, the software removes the extreme bands from the raw HS image\u0026nbsp;(reducing the spectral range from 400-1,000 nm to 400.5-938 nm), as these bands contain significant noise, and then saves the raw HS image. The calibrated HS image is then generated, using Equation\u0026nbsp;(1),\u0026nbsp;and saved on memory as five-digit 16-bit unsigned integers (uint16), where the most significant digit represents the units, and the remaining digits correspond to the decimal places of the transmittance values. Therefore, to obtain true transmittance values, the calibrated HS image must be divided by 10\u003csup\u003e4\u003c/sup\u003e. Additionally, the software creates a synthetic RGB image,\u0026nbsp;following the methodology explained by Ortega \u003cem\u003eet al.\u003csup\u003e33\u003c/sup\u003e\u003c/em\u003e, to facilitate the visualization of the spatial characteristics of the HS image.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePrior to any HS image capture, magnification is selected, in this case the 10\u0026times;. WR and DR reference images are collected.\u0026nbsp;Then, to\u0026nbsp;acquire the HS image, the associated WSI is examined in QuPath\u003csup\u003e35\u003c/sup\u003e to identify an ROI within a specific class, such as IDC (blue), healthy (green) or DCIS (red) tissue. The identified ROI is searched for in the HS microscopic system using the RGB camera and marked down on the histological image using a yellow rectangle (Fig. 2c). The ROI is then captured using the HS microscope, generating the raw HS image, the calibrated HS image, and the synthetic RGB image (Fig. 2d). The RGB image of the ROI is also captured (Fig. 2e) for future analysis. All data corresponding to one of these captures are saved using an identifier with its corresponding metadata, including the patient identifier, classification, and region (e.g., HSI_VNIR_15_IDC_x10_C01; see Data Records section for more details).\u003c/p\u003e\n\u003ch2\u003eData Records\u003c/h2\u003e\n\u003cp\u003eThe HistologyHSI-BC Recurrence Database\u0026nbsp;has been deposited at The Cancer Imaging Archive (TCIA) repository to make it publicly available, organized into a multilayer folder arrangement\u003csup\u003e36\u003c/sup\u003e. The database is divided into three main components: clinical and demographic data, histological WSI and HS images (see Fig. 3a). The clinical and demographic data are stored at the \u003cem\u003e00_01_Clinical_Demographic_Data\u0026nbsp;\u003c/em\u003efile. This XLSX file documents patients\u0026rsquo; demographic status, breast tumor characteristics, treatment received, and their follow-up outcomes (detailed description on Table II).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFurthermore, the histological data are structured in 3 folders. Firstly, \u003cem\u003e01_01_Histological_Images\u003c/em\u003e folder contains the WSI for each patient, stored as MRXS files. Each WSI requires a corresponding metadata folder containing DAT and INI files for proper rendering. Moreover, \u003cem\u003e01_02_Tissue_Annotations\u0026nbsp;\u003c/em\u003efolder includes WSI histological annotations that classify tissue types, with boundaries of the regions outlined in blue (IDC), green (healthy), and red (DCIS), as shown in Fig. 2a. The third folder, \u003cem\u003e01_03_HSI_ROI_Annotations\u003c/em\u003e contains the ROI for each HS image, with the boundaries of the region outlined in yellow (Fig. 2a). Both histological WSI and HS image ROI annotation files are provided in GeoJSON format. A summary of the areas of annotations per patient and tissue type is shown in Table III. Lastly, \u003cem\u003e02_01_HSI_Images\u0026nbsp;\u003c/em\u003efolder contains the HS images of the histological slides, stored in ENVI format\u003csup\u003e37\u003c/sup\u003e Each capture includes the raw HS image, WR and DR calibration files, and the calibrated HS image following the procedure described in Equation (1). As the ENVI standard states, the HS image is saved as a flat-binary raster DAT (data)\u003cem\u003e\u0026nbsp;\u003c/em\u003efile with an accompanying HDR (header)\u003cem\u003e\u0026nbsp;\u003c/em\u003efile containing essential metadata to interpret it. Moreover, within each capture folder a synthetic RGB image and a view of the ROI captured by the RGB camera are stored. The HS image data are stored in folders named according to the regular expression \u003cem\u003eHSI_VNIR_{P}_{TT}_x10_C{CN}\u003c/em\u003e, where {P} represents the patient ID, {TT} indicates the tissue type (IDC, healthy, or DCIS), and {CN} is the capture number (Fig. 3b).\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;III. Summary of histological area coverage and capture counts by tissue type per image.\u003c/p\u003e\n\u003cdiv align=\"center\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"388\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePatient ID\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTissue Annotations Area\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e[mm\u003csup\u003e2\u003c/sup\u003e]\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\" style=\"width: 134px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHS image ROI Annotations\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e[Number of captures]\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 64px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecurrence\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eLabel\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIDC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHealthy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDCIS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIDC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHealthy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDCIS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e15\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e40.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e2.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e64.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e19\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e94.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e24.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e20\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e169.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e58.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e25\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e135.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e139.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e17.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e38\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e210.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e7.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e40\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e72.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e20.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e43\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e57.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e22.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e45\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e71.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e80.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e6.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e47\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e10.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e137.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e51\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e163.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e2.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e2.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e52\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e96.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e3.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e57\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e23.26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e31.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e62\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e4.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e66.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e65\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e104.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e15.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e68\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e21.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e75.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e70\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e55.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e45.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e80\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e4.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e37.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e82\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e109.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e11.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e84\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e88.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e13.20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e7.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e85\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e119.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e5.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e20.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e90\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e179.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e29.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e100\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e68.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e3.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e107\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e8.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e192.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e7.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e112\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e14.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e1.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e124\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e26.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e136\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e190.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e25.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e28.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e138\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e32.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e90.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e11.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e139\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e43.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e72.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e141\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e168.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e7.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e5.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e146\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e24.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e3.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e151\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e3.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e3.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e1.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e152\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e82.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e42.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e19.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e153\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e7.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e77.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e154\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e51.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e162\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e67.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e16.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e1.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e189\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e247.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e197\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e212.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e57.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e4.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e205\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e321.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e35.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e211\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e161.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e213\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e451.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e7.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e229\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e281.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e3.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e2.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e238\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e149.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e255\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e164.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e35.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e259\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e92.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e42.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e2.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e269\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e59.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e6.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e270\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e30.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e111.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e304\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 44px;\"\u003e\n \u003cp\u003e155.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 49px;\"\u003e\n \u003cp\u003e24.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 39px;\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 35px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 41px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003ch2\u003eTechnical Validation\u003c/h2\u003e\n\u003ch2\u003eClinical and Demographic Data Statistic Analysis\u003c/h2\u003e\n\u003cp\u003eA preliminary statistical analysis was conducted to identify differences in the variables between patients with and without recurrence, as shown in\u0026nbsp;Table IV. Statistical tests used for comparisons included the absolute frequency (percentage) for the Chi-square test or Fisher\u0026rsquo;s exact test, and the median (interquartile range) for the Mann-Whitney U test. As expected, certain classic clinical and pathological variables were found to be associated with the risk of developing metastasis in the present cohort\u003csup\u003e8\u003c/sup\u003e, including age at diagnosis, tumor diameter, and LVI.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;IV. Differences in the clinical and demographic variables in recurrence vs. non-recurrence groups.\u0026nbsp;\u003c/p\u003e\n\u003cdiv align=\"center\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"556\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAttribute\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFormat\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecurrence\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNon-Recurrence\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDemographic Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e1: Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e22 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRace\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e1: White\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e22 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthnicity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e1: Hispanic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e22 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMenopausal\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Premenopause\u003c/p\u003e\n \u003cp\u003e1: Postmenopause\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e3 (13.6)\u003c/p\u003e\n \u003cp\u003e19 (86.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e3 (12.0)\u003c/p\u003e\n \u003cp\u003e22 (88.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e1.000*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiagnosis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDx\u0026nbsp;surgery\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Mastectomy\u003c/p\u003e\n \u003cp\u003e1: Lumpectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e7 (31.8)\u003c/p\u003e\n \u003cp\u003e15 (68.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e2 (8.0)\u003c/p\u003e\n \u003cp\u003e23 (92.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.063*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDx\u0026nbsp;age\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eYears\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e73.0 [22.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e57.0 [14.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.017\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor\u0026nbsp;diameter\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eMillimeters\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e26.5 [13.8]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e15.0 [13.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor\u0026nbsp;histologic\u0026nbsp;grade\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e1: Grade 1\u003c/p\u003e\n \u003cp\u003e2: Grade 2\u003c/p\u003e\n \u003cp\u003e3: Grade 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e1 (4.5)\u003c/p\u003e\n \u003cp\u003e11 (50.0)\u003c/p\u003e\n \u003cp\u003e10 (45.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e4 (16.0)\u003c/p\u003e\n \u003cp\u003e14 (56.0)\u003c/p\u003e\n \u003cp\u003e7 (28.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.285*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLVI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Positive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e8 (36.4)\u003c/p\u003e\n \u003cp\u003e14 (63.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e18 (72.0)\u003c/p\u003e\n \u003cp\u003e7 (28.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.031*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePNI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Positive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e15 (68.2)\u003c/p\u003e\n \u003cp\u003e7 (31.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e21 (84.0)\u003c/p\u003e\n \u003cp\u003e4 (16.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.351*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Stage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eT (tumor)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e1: T1\u003c/p\u003e\n \u003cp\u003e2: T2\u003c/p\u003e\n \u003cp\u003e3: T3\u003c/p\u003e\n \u003cp\u003e4: T4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e5 (22.7)\u003c/p\u003e\n \u003cp\u003e14 (63.6)\u003c/p\u003e\n \u003cp\u003e2 (9.1)\u003c/p\u003e\n \u003cp\u003e1 (4.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e17 (68.0)\u003c/p\u003e\n \u003cp\u003e8 (32.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.012*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN (node)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: N0\u003c/p\u003e\n \u003cp\u003e1: N1\u003c/p\u003e\n \u003cp\u003e2: N2\u003c/p\u003e\n \u003cp\u003e3: N3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e8 (36.4)\u003c/p\u003e\n \u003cp\u003e7 (31.8)\u003c/p\u003e\n \u003cp\u003e5 (22.7)\u003c/p\u003e\n \u003cp\u003e2 (9.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eM (metastasis)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: M0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e22 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Molecular Markers\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eER\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Positive ER for \u0026lt; 1% of the TC\u003c/p\u003e\n \u003cp\u003e1: Positive ER for \u0026ge; 1% of the TC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e4 (18.2)\u003c/p\u003e\n \u003cp\u003e18 (81.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e5 (20.0)\u003c/p\u003e\n \u003cp\u003e20 (80.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e1.000*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Positive PR for \u0026lt; 1% of the TC\u003c/p\u003e\n \u003cp\u003e1: Positive PR for \u0026ge; 1% of the TC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e6 (27.3)\u003c/p\u003e\n \u003cp\u003e16 (72.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e9 (36.0)\u003c/p\u003e\n \u003cp\u003e16 (64.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.744*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHER2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Positive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e15 (68.2)\u003c/p\u003e\n \u003cp\u003e7 (31.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e22 (88.0)\u003c/p\u003e\n \u003cp\u003e3 (12.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.154*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eKI67\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: KI67 \u0026lt; 20%\u003c/p\u003e\n \u003cp\u003e1: KI67 \u0026ge; 20%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e4 (18.2)\u003c/p\u003e\n \u003cp\u003e18 (81.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e10 (40.0)\u003c/p\u003e\n \u003cp\u003e15 (60.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.189*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMolecular\u0026nbsp;subtype\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Luminal A\u003c/p\u003e\n \u003cp\u003e1: Luminal B HER2-\u003c/p\u003e\n \u003cp\u003e2: Luminal B HER2+\u003c/p\u003e\n \u003cp\u003e3: HER2+\u003c/p\u003e\n \u003cp\u003e4: Triple negative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e4 (18.2)\u003c/p\u003e\n \u003cp\u003e10 (45.5)\u003c/p\u003e\n \u003cp\u003e4 (18.2)\u003c/p\u003e\n \u003cp\u003e3 (13.6)\u003c/p\u003e\n \u003cp\u003e1 (4.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e6 (24.0)\u003c/p\u003e\n \u003cp\u003e13 (52.0)\u003c/p\u003e\n \u003cp\u003e2 (8.0)\u003c/p\u003e\n \u003cp\u003e1 (4.0)\u003c/p\u003e\n \u003cp\u003e3 (12.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.512*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"7\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLymph Nodes Status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Isolated TC (CC \u0026le; 0.2 mm)\u003c/p\u003e\n \u003cp\u003e2: Micrometastasis (0.2 mm \u0026lt; CC \u0026le; 2 mm)\u003c/p\u003e\n \u003cp\u003e3: Macrometastasis (CC \u0026gt; 2 mm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e6 (27.3)\u003c/p\u003e\n \u003cp\u003e2 (9.1)\u003c/p\u003e\n \u003cp\u003e3 (13.6)\u003c/p\u003e\n \u003cp\u003e11 (50.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e22 (88.0)\u003c/p\u003e\n \u003cp\u003e3 (12.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;ITC\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eNumber of lymph nodes with isolated tumor cells\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e0.0 [0.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e0.0 [0.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.720\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;MICRO\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eNumber of lymph nodes with micrometastasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e0.0 [0.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e0.0 [0.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.027\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;MACRO\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eNumber of lymph nodes with macrometastasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e0.5 [7.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e0.0 [0.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLN\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eTotal number of lymph nodes removed during sentinel lymph node biopsy and/or lymph node dissection\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e13.0 [15.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e2.0 [2.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSLN\u0026nbsp;number\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eNumber of sentinel lymph nodes removed during sentinel lymph node biopsy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e0.5 [2.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e2.0 [2.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSLN\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Negative\u003c/p\u003e\n \u003cp\u003e1: Isolated TC (CC \u0026le; 0.2 mm)\u003c/p\u003e\n \u003cp\u003e2: Micrometastasis (0.2 mm \u0026lt; CC \u0026le; 2 mm)\u003c/p\u003e\n \u003cp\u003e3: Macrometastasis (CC \u0026gt; 2 mm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e4 (36.4)\u003c/p\u003e\n \u003cp\u003e2 (18.2)\u003c/p\u003e\n \u003cp\u003e2 (18.2)\u003c/p\u003e\n \u003cp\u003e3 (27.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e22 (88.0)\u003c/p\u003e\n \u003cp\u003e3 (12.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor Treatment\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;hormonal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e5 (22.7)\u003c/p\u003e\n \u003cp\u003e17 (77.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e4 (16.0)\u003c/p\u003e\n \u003cp\u003e21 (84.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.715*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;CT\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e11 (50.0)\u003c/p\u003e\n \u003cp\u003e11 (50.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e14 (56.0)\u003c/p\u003e\n \u003cp\u003e11 (44.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.906*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;trastuzumab\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e19 (86.4)\u003c/p\u003e\n \u003cp\u003e3 (13.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e23 (92.0)\u003c/p\u003e\n \u003cp\u003e2 (8.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e0.654*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTx\u0026nbsp;RT\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Not received\u003c/p\u003e\n \u003cp\u003e1: Received\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e3 (13.6)\u003c/p\u003e\n \u003cp\u003e19 (86.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e3 (12.0)\u003c/p\u003e\n \u003cp\u003e22 (88.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e1.000*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFollow-up\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMetastasis\u0026nbsp;type\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: No evidence of local or distant metastases\u003c/p\u003e\n \u003cp\u003e1: Metastasized on nearby tissues or lymph nodes\u003c/p\u003e\n \u003cp\u003e2: Metastasized in distant organs from primary site\u003c/p\u003e\n \u003cp\u003e3: Both local and distant metastases are present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e20 (90.9)\u003c/p\u003e\n \u003cp\u003e2 (9.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDFS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eMonths\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e39.0 [48.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e150.0 [28.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVital\u0026nbsp;status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Alive\u003c/p\u003e\n \u003cp\u003e1: Deceased\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e2 (9.1)\u003c/p\u003e\n \u003cp\u003e20 (90.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e21 (84.0)\u003c/p\u003e\n \u003cp\u003e4 (16.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDeath\u0026nbsp;cause\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003e0: Other causes / Still alive\u003c/p\u003e\n \u003cp\u003e1: Cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e5 (22.7)\u003c/p\u003e\n \u003cp\u003e17 (77.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e25 (100.0)\u003c/p\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 226px;\"\u003e\n \u003cp\u003eMonths\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 62px;\"\u003e\n \u003cp\u003e66.5 [85.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e150.0 [28.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 40px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003csup\u003e\u0026Dagger;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eData are expressed as absolute frequency (percentage) for qualitative variables, compared using the Chi-square test or Fisher\u0026rsquo;s exact text*, and as median [interquartile range] for quantitative variables, analyzed using the Mann-Whitney U test\u003csup\u003e\u0026Dagger;\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eCC, cluster cells; CT, chemotherapy; DFS, disease-free survival; Dx, diagnosis; ER, estrogen receptors; HER2, human epidermal growth factor receptor 2; ITC, isolated tumor cells; KI67, proliferation index; LN, lymph node; LVI, lymphovascular invasion; MACRO, macrometastasis; MICRO, micrometastasis; OS, overall survival; PNI, perineural invasion; PR, progesterone receptors; RT, radiotherapy; SLN, sentinel lymph node; TC, tumor cells; Tx, treatment.\u003c/p\u003e\n\u003cp\u003eAnalysis of lymph nodes status revealed a significantly higher percentage of micrometastasis and macrometastasis in patients with recurrence compared to those without recurrence. This association remained significant when considering the number of affected lymph nodes. A similar trend was observed in sentinel lymph nodes status, where micrometastasis and macrometastasis were more prevalent in the recurrence group. However, no significant differences were found in lymph nodes containing isolated tumor cells, classified as negative lymph nodes\u003csup\u003e38\u003c/sup\u003e.\u0026nbsp;These findings align with established knowledge that lymph node metastasis is associated with a higher risk of recurrence in BC patients during follow-up\u003csup\u003e5,39\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eRegarding patient follow-up, we confirmed that all patients without recurrence show no evidence of local or distant metastases, whereas patients with recurrence do, with most of them having metastases only in distant organs and a smaller percentage presenting with both local and distant metastases. Among patients with recurrence, 90.9% died, with cancer being the cause of death in 77.3% of cases. In contrast, among the patients without recurrence who died, none died from cancer. As expected, the median disease-free survival (DFS) was significantly shorter in the recurrence group compared to the non-recurrence group, as was overall survival (OS).\u003c/p\u003e\n\u003ch2\u003eHistopathology WSIs and Annotation Validation\u003c/h2\u003e\n\u003cp\u003ePathologists qualitatively verified the quality of histopathological slides after the sectioning, processing, and staining phases. They confirmed the absence of artifacts in the ROIs of the WSIs, ensuring that these were not introduced during the tissue preparation or digitization phases. Pathologists did not detect the presence of folds, broken tissues, tears, bubbles, scalpel marks, or bad staining on the ROIs due to the tissue preparation phase. Furthermore, they verified the quality of the digitized histopathological slides making sure there were no issues on the WSIs due to the scanning phase. They confirmed the absence of scanning artifacts like focus issues or white reference problems. The annotations on the WSIs (IDC, healthy, and DCIS) were initially made by one pathologist, with the annotation process then subjected to validation by a second pathologist through a pairwise review. This validation phase played a crucial role in minimizing the inter-observability issue, ensuring consistency and reliability in the annotations. An example of the validated annotations is shown in Fig. 4.\u003c/p\u003e\n\u003ch2\u003eHSI Validation\u003c/h2\u003e\n\u003cp\u003eA technical validation was performed to ensure the quality of the HistologyHSI-BC Recurrence Database. The HS microscope employed in this study has been thoroughly characterized in previous works\u003csup\u003e40\u003c/sup\u003e, confirming its strong performance for spectral resolution-intensive applications. The system demonstrates a dynamic range of 65.3 \u0026plusmn; 0.1 dB in transmittance mode, with a constant dark current of 20 digital numbers, which contributes to a reliable HS image capture. It is capable of capturing 826 spectral bands, providing detailed spectral information and accurately reflecting the spectral properties of the materials under analysis. This is evidenced by a spectral correlation measure of 0.88 \u0026plusmn; 0.01 when capturing the WCT-2065 transmittance wavelength calibration standard (Avian Technologies, New London, USA) with a known spectral signature in transmittance mode. The system offers a spatial resolution of 0.739 \u0026plusmn; 0.001 \u0026micro;m/pixel, along with a modulation transfer function (MTF) of 370 \u0026plusmn; 10 line pairs/mm, ensuring sufficient detail for microscopic imaging. Spatial scanning accuracy is indicated by an eccentricity of 0.04 \u0026plusmn; 0.04, and spatial repeatability is shown to have a relative difference of 14 \u0026plusmn; 8% across consecutive captures. All values were measured at 10\u0026times; magnification, the same magnification used for the HS image capturing process in this work.\u003c/p\u003e\n\u003cp\u003eThe characterization parameters obtained from the HS microscopic system demonstrate its capability to provide reliable and accurate HS data. The HS images captured from the 47 patients studied underwent a calibration. Afterwards, the database was evaluated to ensure the quality of the captured data. All HS images from each patient and tissue class (IDC, healthy and DCIS) were averaged for visualization purposes. Fig. 5 groups the spectral signatures of patients with and without recurrence after 12 years. Interestingly, in the biopsies, patients without recurrence showed a greater similarity between healthy and DCIS tissues, while these tissue types were more distinctly separated in patients with recurrence. This finding raises the possibility that the closer resemblance of DCIS to healthy tissue could serve as an indicator of non-recurrence.\u003c/p\u003e\n\u003ch2\u003eUsage Notes\u003c/h2\u003e\n\u003ch2\u003eVisualizing histopathology WSIs\u003c/h2\u003e\n\u003cp\u003eThe authors recommend downloading and installing the QuPath software\u003csup\u003e35\u003c/sup\u003e to visualize and work with the WSIs (MRXS format) and their related annotations (GeoJSON format) (Fig. 2a). There are two ways to open a WSI: drag and drop the MRXS file into QuPath or go on \u0026quot;File/Open\u0026quot; and select and open the MRXS file. There is also a tab on the left side of QuPath\u0026apos;s user interface called \u0026quot;Image\u0026quot;, in which it is possible to analyze the metadata of the histopathological image, such as width, height, magnification, and resolution. After opening the WSI on QuPath, the two available GeoJSON files containing annotations on the WSI should be imported. One includes the annotations related to the tissue compartments (IDC in red, healthy in green, and DCIS in blue). In contrast, the other defines the ROIs used for capturing the HS images, represented as yellow rectangles. These two files can be opened by dragging and dropping them into QuPath or clicking \u0026quot;File/Import objects from file\u0026quot; and selecting the GeoJSON files. The data from the GeoJSON files is visible by clicking on the tab \u0026quot;Annotations\u0026quot;. If the annotation classes are not shown after clicking the \u0026quot;Annotations\u0026quot; tab, click on the button with the three vertical dots on the bottom right of the tab panel, then select \u0026quot;Populate from existing objects/All classes (including sub-classes)\u0026quot; and the class types along with the number of annotations for each will appear.\u003c/p\u003e\n\u003ch2\u003eRecommended histopathology WSI processing\u003c/h2\u003e\n\u003cp\u003eThis section provides guidelines for working with WSIs (MRXS format). Due to their high resolution, efficient processing techniques are necessary to optimize performance and memory usage. Processing high-resolution images can be time-consuming and memory intensive. The highest available resolution of the selected slide image is approximately 85,000 \u0026times; 202,000 pixels, making it significantly large. To optimize performance, a lower resolution (approximately 670 x 1,600) level should be selected for visualization. It is also important to downscale the annotations to match the selected lower resolution level.\u003c/p\u003e\n\u003ch2\u003eRecommended HSI processing\u003c/h2\u003e\n\u003cp\u003eAfter HS data capture, the calibration of HS images is a mandatory step; however, additional processing may be performed depending on the specific application of the data.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eGiven the strong correlation between adjacent spectral bands, spectral dimensionality reduction can be beneficial in reducing intrinsic Gaussian noise and computational costs. This can be accomplished by averaging adjacent spectral bands to create a spectrally reduced HS image. For example, the data could be reduced from the original 826 bands to 275 using a spectral window that includes three neighboring bands.\u003c/li\u003e\n \u003cli\u003eNormalization is also recommended when partial absorbance is less critical, but the specific absorption wavelengths are significant. This normalization can be performed to scale the data between 0 and 1 or to have a mean of 0 and a standard deviation of 1.\u003c/li\u003e\n \u003cli\u003eFor HS analysis of the samples, it is advised to remove the sample background by identifying areas with no absorbance, typically represented by the white background.\u003c/li\u003e\n \u003cli\u003eThe classification of the data can be based on recurrence status and/or tissue type (IDC, healthy, or DCIS). When using ML or DL, it is crucial to ensure that data from the same patient do not appear simultaneously in the training, testing, or validation sets.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2\u003eHow to combine the different databases (Clinical and Demographic, WSIs, and HSI)\u003c/h2\u003e\n\u003cp\u003eIntegrating the diverse databases within the Histology HSI-BC Recurrence Database can enhance the prediction of distant recurrence in BC by leveraging complementary information from multiple modalities. Histopathological WSIs provide morphological insights assessed by pathologists, HS images capture biochemical variations that may indicate early tumor progression, and clinical and demographic data offer critical patient-specific factors. By integrating these databases, researchers can develop more robust predictive models that go beyond traditional histological or clinical assessments, improving risk stratification and supporting personalized treatment decisions.\u003c/p\u003e\n\u003cp\u003eResearchers can employ various data fusion strategies to achieve this integration. \u003cem\u003eEarly fusion\u003c/em\u003e involves combining raw or preprocessed features from each modality before model training, allowing the model to learn directly from the integrated data42. \u003cem\u003eIntermediate fusion\u003c/em\u003e entails extracting high-level features from each database separately and then merging them into a joint representation, capturing modality-specific patterns prior to integration43. \u003cem\u003eLate fusion\u003c/em\u003e consists of training independent models for each modality and subsequently combining their outputs to improve overall prediction accuracy44. Implementing these fusion techniques requires meticulous preprocessing to ensure compatibility and maximize the value of each database. By effectively integrating these multimodal databases, researchers can uncover subtle patterns associated with BC recurrence, advancing precision oncology and personalized patient care.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eCode Availability\u003c/p\u003e\n\u003cp\u003eThe supplementary files related to this study were created using custom Python scripts. Additionally, several widely used toolboxes and libraries can be accessed via the publicly available repository at https://github.com/HIRIS-Lab/HistologyHSI-BC-Recurrence.\u003c/p\u003e\n\u003cp\u003eA tutorial is provided to show how to manipulate annotations in GeoJSON format by overlaying tissue compartments (IDC, healthy, and DCIS) onto the slide image using their corresponding colors: blue, green, and red. Working with WSIs in Python requires the installation of the OpenSlide library. Additionally, the JSON library is needed for loading annotations, while the PIL library is used to prepare and save slide images.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThis work has been supported by the grant Dr. Ferran 2021 (FFPI21/BE01). Moreover, this work was completed while L.Q-Q. was beneficiary of the pre-doctoral grant given by the \u0026ldquo;Agencia Canaria de Investigaci\u0026oacute;n, Innovaci\u0026oacute;n y Sociedad de la Informaci\u0026oacute;n (ACIISI)\u0026rdquo; of the \u0026ldquo;Consejer\u0026iacute;a de Econom\u0026iacute;a, Conocimiento y Empleo\u0026rdquo;, which is part-financed by the European Social Fund (FSE) (POC 2014-2020, Eje 3 Tema Prioritario 74 (85%)). E.S-C. was beneficiary of a grant from the Generalitat de Catalunya (Doctorats Industrials, Pla DI, AGAUR, Grant No. 2022DI057). J.S-N. was supported by the 2024 predoctoral grant from Las Palmas provincial headquarters of the Scientific Foundation of the Spanish Association Against Cancer (PRDLP246561SANT). A.F., a member of the BosomShield project, received funding from Marie Sklodowska-Curie Doctoral Networks Actions (HORIZON-MSCA-2021-DN-01-01) under grant agreement 101073222. C.L-P. was the PI of the SCARLET, a project funded by Proyectos Estrategicos Orientados a la Transici\u0026oacute;n Ecol\u0026oacute;gica y a la Transici\u0026oacute;n Digital, from the 2021 call of the Ministerio de Ciencia e Innovaci\u0026oacute;n, with grant number TED2021-130081B-C22 and funding from NextGenerationEU. Moreover, this work is part of the OASIS (Open Ai-driven Stack para plataformas hpec mejoradas en Sistemas Integrado) project (PID2023-148285OB-C43), supported by the Spanish Government and European Union (FEDER Funds).\u003c/p\u003e\n\u003cp\u003eAuthor contributions\u003c/p\u003e\n\u003cp\u003eL.Q-Q. conceived the study, designed and established the experimental system, carried out a research stay at the Hospital de Tortosa Verge de la Cinta, generated the hyperspectral database, performed image processing and result analysis, and wrote the manuscript. E.S-C. conceived the study, acquired the funding, collected and cleaned the clinical database, generated the histology database, conducted the technical validation experiments, and wrote the manuscript. A.F. generated the histology database and wrote the manuscript. J.S-N. generated the hyperspectral database and wrote the manuscript. S.O. supervised the study and reviewed the manuscript. N.G-B. collected and cleaned the clinical database and reviewed the manuscript. A.F-C. conducted the technical validation experiments and reviewed the manuscript. T.S-A. collected and cleaned the clinical database and reviewed the manuscript. H.F. supervised the study and reviewed the manuscript. L.A-L., D.M-C., and R.B-P. prepared and diagnosed the histological samples, annotated the histopathology images, and reviewed the manuscript. M.L. supervised the study and reviewed the manuscript. G.M.C. and C.L-P. acquired the funding, supervised the study, and reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflict of interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eBray, F. \u003cem\u003eet al.\u003c/em\u003e Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. \u003cem\u003eCA Cancer J Clin\u003c/em\u003e \u003cstrong\u003e74\u003c/strong\u003e, 229\u0026ndash;263 (2024).\u003c/li\u003e\n \u003cli\u003eChiang, A. C. \u0026amp; Massagu\u0026eacute;, J. Molecular Basis of Metastasis. \u003cem\u003eN Engl J Med\u003c/em\u003e \u003cstrong\u003e359\u003c/strong\u003e, 2814 (2008).\u003c/li\u003e\n \u003cli\u003eColzani, E. \u003cem\u003eet al.\u003c/em\u003e Time-dependent risk of developing distant metastasis in breast cancer patients according to treatment, age and tumour characteristics. \u003cem\u003eBr J Cancer\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 1378\u0026ndash;1384 (2014).\u003c/li\u003e\n \u003cli\u003eSiegel, R. L., Giaquinto, A. N. \u0026amp; Jemal, A. Cancer statistics, 2024. \u003cem\u003eCA Cancer J Clin\u003c/em\u003e \u003cstrong\u003e74\u003c/strong\u003e, 12\u0026ndash;49 (2024).\u003c/li\u003e\n \u003cli\u003eRiggio, A. I., Varley, K. E. \u0026amp; Welm, A. L. The lingering mysteries of metastatic recurrence in breast cancer. \u003cem\u003eBr J Cancer\u003c/em\u003e \u003cstrong\u003e124\u003c/strong\u003e, 13\u0026ndash;26 (2021).\u003c/li\u003e\n \u003cli\u003eCaswell-Jin, J. L. \u003cem\u003eet al.\u003c/em\u003e Change in Survival in Metastatic Breast Cancer with Treatment Advances: Meta-Analysis and Systematic Review. \u003cem\u003eJNCI Cancer Spectr\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, (2018).\u003c/li\u003e\n \u003cli\u003eIwata, H. Future treatment strategies for metastatic breast cancer: curable or incurable? \u003cem\u003eBreast Cancer\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 200\u0026ndash;205 (2012).\u003c/li\u003e\n \u003cli\u003eHarbeck, N. \u003cem\u003eet al.\u003c/em\u003e Breast cancer. \u003cem\u003eNature Reviews Disease Primers 2019 5:1\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1\u0026ndash;31 (2019).\u003c/li\u003e\n \u003cli\u003eFasching, P. A. \u003cem\u003eet al.\u003c/em\u003e Biomarkers in Patients with Metastatic Breast Cancer and the PRAEGNANT Study Network. \u003cem\u003eGeburtshilfe Frauenheilkd\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 41 (2015).\u003c/li\u003e\n \u003cli\u003eLee, K. K. \u003cem\u003eet al.\u003c/em\u003e Prognostic Biomarkers for Breast Cancer Metastasis. \u003cem\u003eCancer Metastasis\u003c/em\u003e (2018) doi:10.5772/INTECHOPEN.80576.\u003c/li\u003e\n \u003cli\u003eRoss, J. S. \u003cem\u003eet al.\u003c/em\u003e Immunotherapy predictive biomarkers in metastatic breast cancer (MBC). \u003cem\u003eJournal of Clinical Oncology\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 1023\u0026ndash;1023 (2019).\u003c/li\u003e\n \u003cli\u003ePrasad, S. N. \u0026amp; Houserkova, D. THE ROLE OF VARIOUS MODALITIES IN BREAST IMAGING. \u003cem\u003eBiomedical Papers\u003c/em\u003e \u003cstrong\u003e151\u003c/strong\u003e, 209\u0026ndash;218 (2007).\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eBreast Tumours\u003c/em\u003e. \u003cem\u003eWorld Health Organization Classification of tumours\u003c/em\u003e (IACR, Lyon, France, 2019).\u003c/li\u003e\n \u003cli\u003eJain, E. \u003cem\u003eet al.\u003c/em\u003e Whole Slide Imaging Technology and Its Applications: Current and Emerging Perspectives. \u003cem\u003eInt J Surg Pathol\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 433\u0026ndash;448 (2024).\u003c/li\u003e\n \u003cli\u003ePantanowitz, L. \u003cem\u003eet al.\u003c/em\u003e Twenty Years of Digital Pathology: An Overview of the Road Travelled, What is on the Horizon, and the Emergence of Vendor-Neutral Archives. \u003cem\u003eJ Pathol Inform\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 40 (2018).\u003c/li\u003e\n \u003cli\u003eMezei, T., Kolcs\u0026aacute;r, M., Jo\u0026oacute;, A. \u0026amp; Gurzu, S. Image Analysis in Histopathology and Cytopathology: From Early Days to Current Perspectives. \u003cem\u003eJ Imaging\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 252 (2024).\u003c/li\u003e\n \u003cli\u003eSoliman, A., Li, Z. \u0026amp; Parwani, A. V. Artificial intelligence\u0026rsquo;s impact on breast cancer pathology: a literature review. \u003cem\u003eDiagn Pathol\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 38 (2024).\u003c/li\u003e\n \u003cli\u003eZhao, Y. \u003cem\u003eet al.\u003c/em\u003e A review of cancer data fusion methods based on deep learning. \u003cem\u003eInformation Fusion\u003c/em\u003e \u003cstrong\u003e108\u003c/strong\u003e, 102361 (2024).\u003c/li\u003e\n \u003cli\u003eLlinas-Bertran, A., Butjosa-Esp\u0026iacute;n, M., Barberi, V. \u0026amp; Seoane, J. A. Multimodal data integration in early-stage breast cancer. \u003cem\u003eThe Breast\u003c/em\u003e \u003cstrong\u003e80\u003c/strong\u003e, 103892 (2025).\u003c/li\u003e\n \u003cli\u003eLipkova, J. \u003cem\u003eet al.\u003c/em\u003e Artificial intelligence for multimodal data integration in oncology. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 1095\u0026ndash;1110 (2022).\u003c/li\u003e\n \u003cli\u003eChang, C.-I. \u003cem\u003eHyperspectral Imaging\u003c/em\u003e. (Springer US, Boston, MA, 2003). doi:10.1007/978-1-4419-9170-6.\u003c/li\u003e\n \u003cli\u003eFei, B. Hyperspectral imaging in medical applications. in 523\u0026ndash;565 (2019). doi:10.1016/B978-0-444-63977-6.00021-3.\u003c/li\u003e\n \u003cli\u003eKarim, S., Qadir, A., Farooq, U., Shakir, M. \u0026amp; Laghari, A. A. Hyperspectral Imaging: A Review and Trends towards Medical Imaging. \u003cem\u003eCurr Med Imaging Rev\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 417\u0026ndash;427 (2022).\u003c/li\u003e\n \u003cli\u003eLu, G. \u0026amp; Fei, B. Medical hyperspectral imaging: a review. \u003cem\u003eJ Biomed Opt\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 010901 (2014).\u003c/li\u003e\n \u003cli\u003eSaccomandi, P. \u003cem\u003eet al.\u003c/em\u003e Estimation of anisotropy coefficient of swine pancreas, liver and muscle at 1064 nm based on goniometric technique. \u003cem\u003eJ Biophotonics\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 422\u0026ndash;428 (2015).\u003c/li\u003e\n \u003cli\u003eWang, Q., Wang, J., Zhou, M., Li, Q. \u0026amp; Wang, Y. Spectral-spatial feature-based neural network method for acute lymphoblastic leukemia cell identification via microscopic hyperspectral imaging technology. \u003cem\u003eBiomed Opt Express\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 3017 (2017).\u003c/li\u003e\n \u003cli\u003eEnfield, K. S. S. \u003cem\u003eet al.\u003c/em\u003e Hyperspectral cell sociology reveals spatial tumor-immune cell interactions associated with lung cancer recurrence. \u003cem\u003eJ Immunother Cancer\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 13 (2019).\u003c/li\u003e\n \u003cli\u003eLingle, W. \u003cem\u003eet al.\u003c/em\u003e The Cancer Genome Atlas Breast Invasive Carcinoma Collection (TCGA-BRCA). \u003cem\u003eThe Cancer Imaging Archive\u003c/em\u003e (2016) doi:10.7937/K9/TCIA.2016.AB2NAZRP.\u003c/li\u003e\n \u003cli\u003eCurtis, C. \u003cem\u003eet al.\u003c/em\u003e The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e486\u003c/strong\u003e, 346\u0026ndash;352 (2012).\u003c/li\u003e\n \u003cli\u003eFiorin, A., L\u0026oacute;pez Pablo, C., Lejeune, M., Hamza Siraj, A. \u0026amp; Della Mea, V. Enhancing AI Research for Breast Cancer: A Comprehensive Review of Tumor-Infiltrating Lymphocyte Datasets. \u003cem\u003eJournal of Imaging Informatics in Medicine\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 2996\u0026ndash;3008 (2024).\u003c/li\u003e\n \u003cli\u003eAmgad, M. \u003cem\u003eet al.\u003c/em\u003e Structured crowdsourcing enables convolutional segmentation of histology images. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 3461\u0026ndash;3467 (2019).\u003c/li\u003e\n \u003cli\u003eZhang, Y., Wang, Y., Zhang, B. \u0026amp; Li, Q. A hyperspectral dataset of precancerous lesions in gastric cancer and benchmarks for pathological diagnosis. \u003cem\u003eJ Biophotonics\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, (2022).\u003c/li\u003e\n \u003cli\u003eOrtega, S. \u003cem\u003eet al.\u003c/em\u003e Publisher Correction: Histological Hyperspectral Glioblastoma Dataset (HistologyHSI-GB). \u003cem\u003eSci Data\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 829 (2024).\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eWHO Classification of Tumours of the Breast\u003c/em\u003e. (IARC, Lyon, France, 2012).\u003c/li\u003e\n \u003cli\u003eBankhead, P. \u003cem\u003eet al.\u003c/em\u003e QuPath: Open source software for digital pathology image analysis. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 16878 (2017).\u003c/li\u003e\n \u003cli\u003eQuintana-Quintana, L. \u003cem\u003eet al.\u003c/em\u003e HistologyHSI-BC Recurrence Database. Preprint at (2025).\u003c/li\u003e\n \u003cli\u003eENVI Header Files. https://www.nv5geospatialsoftware.com/docs/ENVIHeaderFiles.html#The.\u003c/li\u003e\n \u003cli\u003eNaidoo, K. \u0026amp; Pinder, S. E. Micro- and macro-metastasis in the axillary lymph node: A review. \u003cem\u003eThe Surgeon\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 76\u0026ndash;82 (2017).\u003c/li\u003e\n \u003cli\u003eJatoi, I., Hilsenbeck, S. G., Clark, G. M. \u0026amp; Osborne, C. K. Significance of Axillary Lymph Node Metastasis in Primary Breast Cancer. \u003cem\u003eJournal of Clinical Oncology\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 2334\u0026ndash;2334 (1999).\u003c/li\u003e\n \u003cli\u003eQuintana-Quintana, L. \u003cem\u003eet al.\u003c/em\u003e Roadmap for the Characterization and Validation of Hyperspectral Microscopic Systems [submitted]. \u003cem\u003eIEEE Trans Instrum Meas\u003c/em\u003e 1\u0026ndash;16 (2025) doi:10.1109/TIM.2024.3379090.\u003c/li\u003e\n \u003cli\u003eWang, Z. \u003cem\u003eet al.\u003c/em\u003e Deep learning-based multi-modal data integration enhancing breast cancer disease-free survival prediction. \u003cem\u003ePrecis Clin Med\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, (2024).\u003c/li\u003e\n \u003cli\u003eNakach, F.-Z., Idri, A. \u0026amp; Goceri, E. A comprehensive investigation of multimodal deep learning fusion strategies for breast cancer classification. \u003cem\u003eArtif Intell Rev\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 327 (2024).\u003c/li\u003e\n \u003cli\u003eLlinas-Bertran, A., Butjosa-Esp\u0026iacute;n, M., Barberi, V. \u0026amp; Seoane, J. A. Multimodal data integration in early-stage breast cancer. \u003cem\u003eThe Breast\u003c/em\u003e \u003cstrong\u003e80\u003c/strong\u003e, 103892 (2025).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Las Palmas de Gran Canaria","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7242335/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7242335/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMetastasis occurs in nearly 1 out of 3 breast cancer (BC) patients and significantly reduces survival rates, particularly in cases of distant metastases. As most distant metastases develop after diagnosis (i.e., recurrence) and remain incurable, there is a critical need for prognostic biomarkers to assess recurrence risk. Multimodal data analysis has emerged as a promising approach to integrate diverse information, offering a more comprehensive perspective. This study introduces the Histology HSI-BC (hyperspectral imaging - breast cancer) Recurrence Database, the first publicly accessible multimodal database designed to advance BC distant recurrence prediction. The database comprises 47 histopathological whole-slide images, 677 hyperspectral (HS) images, and clinical and demographic data from 47 BC patients, of whom 22 (47%) experienced distant recurrence over a 12-year follow-up. Histopathological slides were digitized using a whole-slide scanner and annotated by expert pathologists, while HS images were acquired with an HS camera coupled to a bright-field microscope. This database provides a promising resource for studying BC recurrence prediction and personalized treatment strategies by integrating the aforementioned multimodal data.\u003c/p\u003e","manuscriptTitle":"Histological Hyperspectral Breast Cancer Recurrence Database (HistologyHSI-BC Recurrence)","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-30 08:25:15","doi":"10.21203/rs.3.rs-7242335/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e3b8145f-7bb7-4ae4-a559-0eecdcccdf7d","owner":[],"postedDate":"July 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":52307935,"name":"Biotechnology and Bioengineering"}],"tags":[],"updatedAt":"2025-07-30T08:25:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-30 08:25:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7242335","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7242335","identity":"rs-7242335","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}