Development and Validation of a Framework for Registration of Whole-Mount Radical Prostatectomy Histopathology with Three-dimensional Transrectal Ultrasound

Development and Validation of a Framework for Registration of Whole-Mount Radical Prostatectomy Histopathology with Three-dimensional Transrectal Ultrasound | 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 Development and Validation of a Framework for Registration of Whole-Mount Radical Prostatectomy Histopathology with Three-dimensional Transrectal Ultrasound Auke Jager, Marije J. Zwart, Arnoud W. Postema, Daniel L. Kroonenberg, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5217620/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Apr, 2025 Read the published version in BMC Urology → Version 1 posted 3 You are reading this latest preprint version Abstract Purpose Artificial intelligence (AI) has the potential to improve diagnostic imaging on multiple levels. To develop and validate these AI-assisted modalities a reliable dataset is of utmost importance. The registration of imaging to pathology is an essential step in creating such a dataset. This study presents a comprehensive framework for the registration of 3D transrectal ultrasound (TRUS) to radical prostatectomy specimen (RPS) pathology. Methods The study enrolled patients who underwent 3D TRUS and were scheduled for radical prostatectomy. A four-step process for registering RPS to TRUS was used: image segmentation, 3D reconstruction of RPS pathology, registration and ground truth calculation. Accuracy was assessed using a target-registration error (TRE) based on landmarks visible on both TRUS and pathology. Results 20 Sets of 3D TRUS and RPS pathology were included for analyses. The mean TRE was 3.5 mm, ranging from 0.4 to 5.4 mm. Conclusion The framework proposed in this study accomplishes precise registration between prostate pathology and imaging. The methodologies employed hold the potential for broader application across diverse imaging modalities and other target organs. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Imaging plays an increasingly important role in prostate cancer (PCa) diagnostics. Prostate magnetic resonance imaging (MRI) prior to prostate biopsy is currently the standard of care for patients with a clinical suspicion for PCa (1). While MRI has greatly improved the diagnostic pathway for PCa, it still has certain limitations related to diagnostic accuracy, availability, cost, and interobserver agreement (2, 3). As a result, various methods to improve prostate imaging are being investigated. These methods are related to advancements in MRI performance and efficiency, as well as the exploration of alternative imaging modalities such as transrectal ultrasound (TRUS) and prostate-specific membrane antigen (PSMA) positron emission tomography (PET)-based modalities (4–6). An area of particular interest is the integration of artificial intelligence (AI) into the diagnostic process, which shows promising potential in improving diagnostic accuracy, reducing interobserver variability, and addressing time-intensity concerns (7–9) Regarding TRUS as an alternative modality, the multiparametric approach shows promise. This approach combines information from different ultrasound modalities, including Brightness-mode (B-mode), contrast-enhanced ultrasound (CEUS), and shear wave elastography (SWE). Machine learning (ML) techniques have emerged as a valuable tool for integrating and extracting information from these ultrasound modalities (5, 8, 10). In recent developments, the availability of 3D TRUS has provided further support to these advancements. The capability of imaging the entire prostate gland in a single scan offers evident advantages for the clinical workflow (11–13). (3D) TRUS-based imaging may be advantageous over MRI due to its cost-effectiveness and wider availability, although its value in clinical practice for PCa diagnosis is not yet established (5). Ongoing research aims to develop AI methods to detect PCa on different imaging modalities (11). For development and validation of such methods, a reliable ground truth is required (14). The best way to obtain this ground truth is through correlation of pre-surgical in-vivo imaging with pathology slides obtained from radical prostatectomy specimens (RPS) (15). The pathology slides contain information on the localization and characterization of cancerous tissue, which can then be mapped onto the corresponding images recorded in-vivo. However, correlation of pathology to imaging (i.e., registration) poses several challenges. Following surgery, the RPS is fixated in formalin and sliced into slices of approximately equal size, which are then used to create pathology slides for evaluation by a pathologist. Due to variation in slice thickness, differences between pathology and imaging slice orientation, and RPS deformation, pathology and imaging slices cannot simply be overlaid for correlation (16). For MRI, multiple registration approaches using open-source ML algorithms have been proposed that reach fairly high levels of accuracy (15, 17). These approaches cannot be applied to TRUS due to a different slice orientation and prostate deformation due to the pressure applied by the ultrasound probe on the prostate (16). Previous work has shown promising results using affine and elastic, surface-based registration for in-vivo registration of 2D TRUS with RPS pathology, with a mean target registration error (TRE) of 2.1 mm (18). However, the validation was limited to a small cohort of seven patients. In contrast, the current study employs a larger dataset and uses 3D TRUS, providing a more robust validation. This paper presents a new framework for the registration of 3D TRUS with 3D RPS pathology and reports on its in-vivo validation. Although the framework is primarily developed for TRUS, it proposes several innovative techniques that can potentially facilitate accurate registration of pathology across different imaging modalities. Methods The cohort consisted of patients planned for radical prostatectomy (RP) because of biopsy-proven PCa, prospectively included in a multicentre study. The trial protocol for this study has been published (11). The framework for TRUS-pathology registration can be separated in four steps 1) TRUS acquisition and segmentation 2) RPS pathology handling and 3D reconstruction, 3) TRUS-pathology registration and 4) calculation of ground truth (Fig. 1 ) 1. Three-dimensional transrectal prostate ultrasound acquisition and segmentation All patients underwent 3D multiparametric transrectal ultrasound (3D mpUS) prior to RP using a LOGIQ TM E10 ultrasound machine with a RIC 5-9D 3D endocavitary probe (GE Healthcare, Chicago, IL, USA). A motorized mechanism inside of the probe supports the automatic acquisition of a 120° volumetric sweep. 3D mpUS acquisition consisted of 3D B-mode, 3D shear-wave elastography, and 4D contrast enhanced imaging and is described in detail in the published trial protocol (Fig 2) (11). To ensure stable image acquisition, a fixture device was developed to fixate the ultrasound probe after satisfactory positioning (Fig 3). Image processing consisted of manual segmentation of anatomical structures, using a segmentation tool developed specifically for this study by Angiogenesis Analytics (AA, ‘s-Hertogenbosch, the Netherlands). Segmentations were performed by AJ or AP, with respectively 4 and 8 years of experience with TRUS. Standard segmentation consisted of the prostate border, the border between peripheral zone (PZ) and transitional zone (TZ), and the urethra. Additional segmentation of other anatomical structures (utriculus and ejaculatory ducts) and landmarks (e.g., calcifications and cysts) were only performed when clearly visible on both ultrasound and pathology. 2. Radical prostatectomy pathology 2a. Radical prostatectomy specimen handling RPS handling was performed according to a study specific protocol (11). To facilitate accurate 3D pathology reconstruction and subsequent registration to TRUS, certain steps were included in the protocol. Prior to fixation in formalin, four intravenous cannulas were inserted in four different quadrants of the RPS from apex to base. After at least 24h of fixation in formalin, the cannulas were removed, the RPS was cut into 4-mm thick slices from apex to base using the TruSlice specimen cut up system (CellPath ltd, Newtown, UK) and a photo of the gross pathology slices was captured. The most apical and basal slice were further divided parasagittally in 4-mm strips. Pathology slides were cut at 4-µm thickness from the gross pathology slices using a microtome and scanned on high resolution (40x enlargement, 20x objective, 2.1 camera lens) with a Pannoramic 1000 Digital Slide Scanner (3DHISTECH, H-1141 Budapest, Hungary). Digital pathology slices were annotated according to study protocol by a pathologist using a web-based pathology annotation tool (Slidescore, Amsterdam, the Netherlands)(11). Apart from cancerous areas, the annotation protocol also included the same anatomical structures and landmarks segmented on TRUS. 2b. Three-dimensional pathology reconstruction After annotation, pathology slides were overlaid on the gross image of prostate slices. The gross prostate slices with overlay were then stacked to create a 3D pathology model. The needle holes resulting from the intravenous cannulas were used to prevent rotational and translational errors between the slices. Shrinkage of histopathology slides could be compensated for up to 10%. The apical pathology slices were sequenced from left to right, checked for anterior and posterior orientation, and fitted on the most apical gross prostate slice. The parasagittal pathology slides from the most basal (or bladder neck) slice were not included in the reconstruction. 3. Registration of pathology to ultrasound images The registration is defined by both a rigid and a non-rigid transform. By applying first, the rigid, and then the non-rigid transform, the pathology slides are mapped onto the TRUS images, minimising the distance between the visible anatomical structures in both domains. The registration procedure compensates for in-vivo deformation of the prostate caused by pressure from the ultrasound probe, as well as ex-vivo deformation resulting from the excision of the prostate and subsequent fixation in formalin (19). The full registration algorithm was implemented and visualised in a registration tool (developed by AA), with the individual steps indicated in Fig 4. 3a. Compensation for Probe Deformation During TRUS acquisition, the prostate is deformed in-vivo due to pressure applied by the probe. As this is not present in the pathology slices, compensation of this deformation is required for accurate registration. This is done by modelling the deformation effect of the probe on homogeneous tissue based on a 3D scan of the probe. The inverse of this deformation is then applied to the 3D ultrasound segmentations. 3b. Rigid Registration The first step of the registration is to define a rigid transform, described by a translation, a rotation and three scaling factors. The combination of translation and rotation parameters positions the full reconstructed prostate to the ultrasound scan domain. Two of the scaling factors effectively act as gross and apex slice thicknesses, and the third is applied to all slices in their respective surface planes to account for the aforementioned shrinking factors. The first step in defining this rigid transform, is to align the centre of the ultrasound-scan boundary segmentation with the centre of the boundary annotations from the pathology slices as well as aligning the left-right, anterior-posterior, and apex-base directions in both pathology and scan domain. Subsequently, the parameters are optimized by an iterative closest point (ICP) algorithm. For each visible anatomical structure, a set of point-pairs is defined by pairing all points of the annotations of the respective structure in the pathology domain with the closest point on the segmentations in the scan domain. A loss term for each anatomical structure is then calculated as the root mean square error (RMSE) of the distances between the points in each pair. The total loss is now defined as the weighted sum of the individual loss terms of each set. The transform with minimum total loss is found through gradient-descent and applied to the points in the pathology domain. Subsequently, a new set of closest points is found, and the process repeated until the minimised total loss has converged. The blue ‘tension’ lines in the registration section of Fig. 1 present a visual representation of the point-pairs. As mentioned, the parasagittal pathology slides from the most basal gross slice are not included in the reconstruction. This leads to a significant degree of freedom in the placement of the gross slices over the apex-base direction. As this degree of freedom is poorly reflected in the previously described loss terms, an additional loss term is included to reflect the difference between a desired base height and the effective base height resulting from the current rigid transform parameters. The registration tool provides for means for an operator to monitor the registration process and manually adjust some of the key parameters. These adjustable parameters include weight factors of the described loss terms, a desired base height and adjustment for shear. The weight factors of the anatomical structures are adjusted to aim for a homogeneous distribution of the loss over each structure. The desired base height is adjusted based on visual inspection of tension lines, aiming to settle at a gross slice thickness of to account for the shrinking mentioned earlier. Any indication of shear in the stack of pathology slices is also manually adjusted for. The optimization iterations are terminated once the total loss has converged, and the operator recognises no need for further adjustments. To account for the operational variation, this step was reviewed by a 2nd operator and discussed where necessary to reach consensus. 3c. Non-Rigid Registration The second step of the registration is to define a non-rigid transform which is applied after the rigid transform. The non-rigid transform compensates for local deformation, assuming homogeneous elasticity of the tissue. Following the concept of the Finite Element Method (FEM), the scan space is subdivided into a regular grid with a size of 2mm. The deformation (i.e. variation in distance) between adjacent grid points is calculated, resulting in one overall additional loss term for elasticity. Finding the deformation that minimizes the elasticity loss term is like mimicking the behaviour of elastic tissue that offers some degree of resistance against deformation. The non-rigid registration phase has now become similar to the rigid registration phase, with the elasticity loss that is added the overall weight with an appropriate weight as the key difference. Iterations of the optimization algorithm are carried out until either a) the difference in loss between one iteration and the next was smaller than a pre-defined value or b) a maximum number of iterations was reached. Note that no input of the operator is required in this step. 4. Calculation of ground truth Applying both the rigid and non-rigid transforms, the pathology annotations are mapped onto the 3D ultrasound scan space which was compensated for probe deformation. Therefore, the last step consists of applying the probe deformation to the annotations. Voxels lying inside the registered slides can now be assigned as either malignant or benign, based on the pathology annotations. Note that this does not provide any information on the voxels lying in between the pathology slides. Interpolation techniques are applied to the voxels lying in between the pathology slides to assign each of these with a probability of their being malignant. Patient selection Patients who exhibited a clearly distinguishable landmark in both the pathology and ultrasound scan domains were selected for the assessment of registration accuracy. The determination of the validity of these landmarks was made through consensus between two independent observers (AJ & AP). Registration accuracy In order to calculate the target-registration error (TRE), the registration was performed with deliberate exclusion of the landmark as an input to the optimization algorithm, thereby mitigating its potential influence on the registration outcome. The landmarks could then be used to calculate the TRE between pathology and ultrasound. Following registration, each landmark annotation on the pathology slides results in a group of registered landmark voxels in ultrasound scan domain, of which the outer points were extracted. Subsequently, an ICP algorithm, similar to the one used in the registration process, was employed to obtain a translation that minimises the difference between the transformed outer voxel points and points on the landmark contour(s) obtained from the scan. To establish point pairs, the untransformed outer voxel point was paired with the point on the landmark contour lying closest to the translated outer voxel point. The TRE was then defined as the mean of the Euclidean distances between these point pairs. The magnitude of the TRE in the apex-base (AB), left-right (LR) and posterior-anterior (PA) axes only was also evaluated by projecting the distances onto each axis. The TRE was determined for each patient, and the overall registration accuracy was subsequently expressed as the mean of these TREs. Results The total cohort consisted of 126 eligible patients, of which 20 exhibited a clearly discernable landmark and were included in the registration accuracy analysis. The overall registration accuracy (mean TRE) was 3.5 mm (range 0.4 mm – 5.4 mm), with a TRE in the AB, LR and PA directions of 2.5, 1.1 and 1.4 respectively. Table 1 gives an overview of the per patient TRE and types of landmarks. A visualization of all ultrasounds, pathology slides and registrations, including their landmarks is provided in supplementary materials 1. Table 1 Per patient TRE. TRE = target registration error, AB = apex-base, LR = left-right, PA = posterior-anterior, BPH = benign prostatic hyperplasia. Prostate volume (mL) Landmark TRE TRE (AB) TRE (LR) TRE (PA) 1 133.06 Cyst 5.2 4.2 1.0 2.1 2 28.39 Calcification 1.9 1.5 0.2 0.1 3 33.46 Calcification 4.2 3.1 1.3 2.2 4 59.08 BPH nodule 2.5 1.3 0.8 1.8 5 88.54 Calcification 5.4 4.4 0.5 2.4 6 35.72 Calcification 5.3 3.6 2.9 2.0 7 80.15 Cyst 2.8 2.2 0.9 0.4 8 45.93 Calcification 3.8 2.6 1.4 2.0 9 45.97 Cyst 0.4 0.1 0.0 0.3 10 90.11 Cyst 4.3 2.5 1.6 3.0 11 76.87 Cyst 2.3 0.6 2.1 0.3 12 28.36 Cyst 3.1 2.6 0.7 0.8 13 62.10 Calcification 1.6 0.5 1.4 0.3 14 100.10 BPH Nodule 4.6 3.4 0.4 3.2 15 47.43 Calcification 4.3 3.5 0.6 1.1 16 28.75 Calcification 1.7 1.2 0.9 0.9 17 25.35 Cyst 5.1 3.8 2.0 3.4 18 47.70 Cyst 3.9 3.5 0.7 0.1 19 28.62 Calcification 4.9 3.9 1.5 1.4 20 27.29 Cyst 2.4 1.0 1.9 0.9 Mean 57.92 - 3.5 2.5 1.1 1.4 Median 46.7 3.9 2.6 0.9 1.3 Discussion The registration of RPS pathology to prostate imaging poses a considerable challenge, particularly in the context of TRUS. The current study presents a comprehensive framework for accurate registration of RPS to 3D TRUS, with a mean TRE of 3.5 mm based on 20 patients with clearly discernable landmarks. While there is no consensus on the maximum acceptable registration error, clinically significant PCa lesions are generally considered to have a volume of ≥ 0.5 cm 3 , which corresponds to a diameter of 10 mm, assuming to lesion is spherical (16). Therefore, we should aim to have a registration error below 10 mm. This study makes significant advancements compared to existing literature in several areas. Prior research has demonstrated the feasibility of registering pathology to ultrasound with acceptable levels of accuracy (18). Our study builds upon this foundation by employing a larger dataset and a more comprehensive methodology. Notably, our study represents the first use of 3D acquired TRUS imaging, as opposed to relying on a 3D reconstruction from handheld 2D transversal sweeps. To ensure precise 3D RPS reconstruction, we implemented needle placement prior to RPS fixation, effectively preventing transitional and rotational errors. Furthermore, to aid the registration, we included compensation for the probe deformation. Although several assumptions and simplifications had to be made to model the deformation, it has shown to be a valuable addition to the registration process. Furthermore, while the use of anatomical structures for improving registration accuracy has generally been limited to the prostate border and urethra (15, 16, 18). Based on our experience the addition of the border between PZ and TZ is feasible for most patients and is highly useful during registration. Although the visibility of ejaculatory ducts can vary (7 out of 20 cases in our dataset), when available, they offer valuable information during registration. When using ultrasound as the imaging modality, it is important to note that these ducts are generally only visible on CEUS and not on conventional B-mode. Additionally, landmarks had to be excluded from the registration to be able to calculate accuracy, but when included, they can provide information on slice and base thickness, and thus have a potential to improve accuracy. In our set of 20 patients, we had 3 patients with 2 landmarks available. For these patients, we repeated the registration but this time including the landmark that had not been used for validation. The new results and their differences compared to the old results are shown in table 2. Table 2 This table illustrates the effect of landmarks on the TRE. The values within the parentheses indicate change in TRE resulting from the integration of a landmark into the registration procedure. TRE = target registration error, AB = apex-base, LR = left-right, PA = posterior-anterior, BPH = benign prostatic hyperplasia. TRE TRE (AB) TRE (LR) TRE (PA) 1 4.9 (-0.6) 3.6 (-0.8) 0.6 (+ 0.1) 2.6 (+ 0.2) 2 2.7 (-1.2) 2.3 (-1.2) 0.6 (-0.1) 0.4 (+ 0.2) 3 2.3 (-2.3) 0.9 (-2.5) 0.7 (+ 0.3) 1.2 (-2.0) This is a very small set of patients and therefore caution should be taken with any conclusions. However, all patients showed a significant overall improvement by inclusion of a landmark in the registration, especially in the apex-base direction. Lastly, our study demonstrates the feasibility of including parasagittally cut apical slides in the pathology reconstruction, which has not been previously described due to the complexities involved in reconstruction (16). These slides offer valuable information, as they often contain PCa lesions (20). In fact, in our dataset, 75% of patients (15 out of 20) exhibited some form of prostate cancer in the apical pathology slides, underscoring the importance of including these slides in a ground truth dataset. To optimize registration accuracy, several actions were taken during the handling of prostate specimens. A commercially available cutting device was employed to ensure uniform thickness of pathology slices, and needle placement prior to fixation facilitated proper stacking of the slides after slicing. The biggest uncertainty in the registration is in the apex-base axis, as we have the least information in that direction. To add further complexity, since the length of the prostate from apex to base may not be evenly divisible by 4, and the apex is cut from apex to base, the thickness of the most basal slide often differs from the other slices. Although the inclusion of the apical slides aids in appropriately positioning the whole-mount slices from apex to base towards the apex, the most basal slides (i.e., bladder neck) were not included in the reconstruction process. The decision to exclude these slides stems from several factors. Firstly, the basal slice is typically larger than the apical slice and is cut thicker to facilitate the evaluation of extracapsular extension. Consequently, pathology slides originating from the basal slice are numerous and often need to be further divided to fit into pathology cassettes. Moreover, the identification of the prostate border becomes more challenging in the vicinity of the urethra and seminal vesicles, further complicating the reconstruction of the basal slide. Additionally, it is worth mentioning that the incidence of prostate cancer in the base is generally low, further justifying the exclusion of the basal slide from the reconstruction process (21). (22). However, this leaves a significant range of possible values for the slice and base thicknesses, with the final chosen value dependent on the operator. This is reflected in our TRE calculations, which show that the TRE in the apex-base direction is twice as large as in the left-right and posterior-anterior directions. Several limitations should be acknowledged in this study. Firstly, the manual segmentation, annotation and registration involved are time intensive, costing approximately an hour per prostate. Efforts to automate segmentation of ultrasound images through deep learning algorithms have shown promising results and are expected to reduce the required manual labor. Moreover, the manual nature of these procedures introduces the possibility of interobserver variability. Certain ultrasound segmentations, such as those for the urethra and zones, may not always be clearly visible, particularly in cases involving larger prostates. Although consensus meetings were held among operators to address uncertainties in specific segmentations, it is important to acknowledge that this process may have influenced the registration accuracy to some extent. A suggested improvement to the registration algorithm would be to provide it with a weighting factor for each segmentation point based on the quality of the scan at that location. This would allow the registration to put more focus on more clearly visible – and therefore more certain - areas of the US segmentations. Lastly, the registration method employed in this study requires custom-made software for execution, which may present challenges when implementing it elsewhere. However, the authors can be contacted for assistance in research projects, and they are willing to share data upon reasonable request, facilitating further exploration and collaboration in the field. Conclusion The study presents a comprehensive framework for registering RPS pathology to 3D TRUS, achieving accurate results based on discernable landmarks. Precise registration is crucial for generating reliable datasets, essential in developing and validating new diagnostic tools and for training AI-assisted imaging modalities. Declarations Conflict of Interest: HB is chair of the clinical board for AA. AP and MM are scientific advisors for AA for which they receive compensation. MZ and WZ are employees of AA. Ethics Approval and Consent to Participate : this study was approved by an accredited medical research ethics committee (MEC AMC) under reference number 2020_268#B202178. All study participants signed an informed consent form that includes the consent for use of their data for publication. The study was performed in accordance with the Declaration of Helsinki. Author Contributions: AJ: Conceptualization, Methodology, Acquisition, Writing – Original Draft, Visualization, Project administration. MZ: conceptualization, Methodology, Writing – Original Draft, Visualization, Project administration, interpretation of data, statistical analysis. DK & AP: Conceptualization, Methodology, Acquisition, Writing – Review & Editing. WZ: Methodology, interpretation of data, statistical analysis. JO: Writing – Review & Editing, Supervision, Funding acquisition. HB: Writing – Review & Editing, Supervision, Funding acquisition. MM: Conceptualization, Methodology, Writing – Review & Editing, Supervision, Project administration. All authors read and approved the final paper. Funding: This study was co-funded by the European Union and Angiogenesis Analytics. Clinical trial number : not applicable. Data Availability Statement: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. References Mottet N, van den Bergh RCN, Briers E, Van den Broeck T, Cumberbatch MG, De Santis M, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol. 2021;79(2):243-62. Drost FH, Osses D, Nieboer D, Bangma CH, Steyerberg EW, Roobol MJ, et al. Prostate Magnetic Resonance Imaging, with or Without Magnetic Resonance Imaging-targeted Biopsy, and Systematic Biopsy for Detecting Prostate Cancer: A Cochrane Systematic Review and Meta-analysis. Eur Urol. 2020;77(1):78-94. Brembilla G, Dell'Oglio P, Stabile A, Damascelli A, Brunetti L, Ravelli S, et al. Interreader variability in prostate MRI reporting using Prostate Imaging Reporting and Data System version 2.1. Eur Radiol. 2020;30(6):3383-92. Lisney AR, Leitsmann C, Strauss A, Meller B, Bucerius JA, Sahlmann CO. The Role of PSMA PET/CT in the Primary Diagnosis and Follow-Up of Prostate Cancer-A Practical Clinical Review. Cancers (Basel). 2022;14(15). Mannaerts CK, Engelbrecht MRW, Postema AW, van Kollenburg RAA, Hoeks CMA, Savci-Heijink CD, et al. Detection of clinically significant prostate cancer in biopsy-naive men: direct comparison of systematic biopsy, multiparametric MRI- and contrast-ultrasound-dispersion imaging-targeted biopsy. BJU Int. 2020;126(4):481-93. Turkbey B, Haider MA. Deep learning-based artificial intelligence applications in prostate MRI: brief summary. Br J Radiol. 2022;95(1131):20210563. de Rooij M, van Poppel H, Barentsz JO. Risk Stratification and Artificial Intelligence in Early Magnetic Resonance Imaging-based Detection of Prostate Cancer. Eur Urol Focus. 2021. Wildeboer RR, Mannaerts CK, van Sloun RJG, Budaus L, Tilki D, Wijkstra H, et al. Automated multiparametric localization of prostate cancer based on B-mode, shear-wave elastography, and contrast-enhanced ultrasound radiomics. Eur Radiol. 2020;30(2):806-15. Yi Z, Hu S, Lin X, Zou Q, Zou M, Zhang Z, et al. Machine learning-based prediction of invisible intraprostatic prostate cancer lesions on (68) Ga-PSMA-11 PET/CT in patients with primary prostate cancer. Eur J Nucl Med Mol Imaging. 2022;49(5):1523-34. Suarez-Ibarrola R, Sigle A, Eklund M, Eberli D, Miernik A, Benndorf M, et al. Artificial Intelligence in Magnetic Resonance Imaging-based Prostate Cancer Diagnosis: Where Do We Stand in 2021? Eur Urol Focus. 2021. Jager A, Postema AW, Mischi M, Wijkstra H, Beerlage HP, Oddens JR. Clinical Trial Protocol: Developing an Image Classification Algorithm for Prostate Cancer Diagnosis on Three-dimensional Multiparametric Transrectal Ultrasound. Eur Urol Open Sci. 2023;49:32-43. Wildeboer RR, van Sloun RJG, Huang P, Wijkstra H, Mischi M. 3-D Multi-parametric Contrast-Enhanced Ultrasound for the Prediction of Prostate Cancer. Ultrasound Med Biol. 2019;45(10):2713-24. Wildeboer RR, Van Sloun RJG, Schalk SG, Mannaerts CK, Van Der Linden JC, Huang P, et al. Convective-Dispersion Modeling in 3D Contrast-Ultrasound Imaging for the Localization of Prostate Cancer. IEEE Trans Med Imaging. 2018;37(12):2593-602. Willemink MJ, Koszek WA, Hardell C, Wu J, Fleischmann D, Harvey H, et al. Preparing Medical Imaging Data for Machine Learning. Radiology. 2020;295(1):4-15. Rusu M, Shao W, Kunder CA, Wang JB, Soerensen SJC, Teslovich NC, et al. Registration of presurgical MRI and histopathology images from radical prostatectomy via RAPSODI. Med Phys. 2020;47(9):4177-88. Wildeboer RR, van Sloun RJG, Postema AW, Mannaerts CK, Gayet M, Beerlage HP, et al. Accurate validation of ultrasound imaging of prostate cancer: a review of challenges in registration of imaging and histopathology. J Ultrasound. 2018;21(3):197-207. Shao W, Banh L, Kunder CA, Fan RE, Soerensen SJC, Wang JB, et al. ProsRegNet: A deep learning framework for registration of MRI and histopathology images of the prostate. Med Image Anal. 2021;68:101919. Schalk SG, Postema A, Saidov TA, Demi L, Smeenge M, de la Rosette JJ, et al. 3D surface-based registration of ultrasound and histology in prostate cancer imaging. Comput Med Imaging Graph. 2016;47:29-39. Jonmarker S, Valdman A, Lindberg A, Hellstrom M, Egevad L. Tissue shrinkage after fixation with formalin injection of prostatectomy specimens. Virchows Arch. 2006;449(3):297-301. Veerman H, Boellaard TN, van Leeuwen PJ, Vis AN, Bekers E, Hoeks C, et al. The detection rate of apical tumour involvement on preoperative MRI and its impact on clinical outcomes in patients with localized prostate cancer. J Robot Surg. 2022;16(5):1047-56. Vargas SO, Jiroutek M, D'Amico AV, Renshaw AA. Distribution of carcinoma in radical prostatectomy specimens in the era of serum prostate-specific antigen testing. Implications for delivery of localized therapy. Am J Clin Pathol. 1999;112(3):373-6. Sakamoto Y, Fukaya K, Haraoka M, Kitamura K, Toyonaga Y, Tanaka M, et al. Analysis of prostate cancer localization toward improved diagnostic accuracy of transperineal prostate biopsy. Prostate Int. 2014;2(3):114-20. Additional Declarations Competing interest reported. Conflict of Interest: HB is chair of the clinical board for AA. AP and MM are scientific advisors for AA for which they receive compensation. MZ and WZ are employees of AA. Supplementary Files Supplementarymaterials1.pdf Supplementary Material Supplementary materials 1: A visualization the ground truth for all 20 included patients. The 3D segmentations the ultrasound scans are visualized in red and green, red being the right prostate lobe and green the left. Within the prostate in yellow the border between the TZ and PZ is visualized, as well as the urethra and the spermatic ducts. The pathology slides are mapped onto the segmented ultrasound scans, the borders of which are shown as purple dots and blue lines. Cite Share Download PDF Status: Published Journal Publication published 03 Apr, 2025 Read the published version in BMC Urology → Version 1 posted Editor assigned by journal 07 Nov, 2024 Submission checks completed at journal 31 Oct, 2024 First submitted to journal 07 Oct, 2024 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-5217620","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":375916101,"identity":"64756856-c7d7-43d5-bcea-82b74777461b","order_by":0,"name":"Auke Jager","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABA0lEQVRIiWNgGAWjYHCCBBDBww/lMTZAGTKEtMhIgpQeQNLCQ8gqG4MDxGrhb2B4+PFLxT0e4+Onkz9/qLgn2z/tjOmGHwx3cGqROMCQLC1zppjH7EzuNokDZ4qNZ9zOMbvZw/AMt8MOMCRIS7Yl8JgdyN3GcLAtIbEBqOU2A8NhnFrkgbb8lvyXwGPc/3bzh4P/EhLnE9IC9HWa5MeGBB4DidwNEgcbEhI3ENJieJghzZrhWAKPxI232yTOHEsw3ng7rexmjwFuLXLHe5Jv/qhJsOfvz938oaImQXbe7eRtN35UHJbD6X1mngRmLAYa4NQABOwHGH/gkx8Fo2AUjIJRAADsaV/qM6MX9wAAAABJRU5ErkJggg==","orcid":"","institution":"Amsterdam UMC, University of Amsterdam","correspondingAuthor":true,"prefix":"","firstName":"Auke","middleName":"","lastName":"Jager","suffix":""},{"id":375916102,"identity":"4d9fb00b-b586-4327-9671-2684eaa111f0","order_by":1,"name":"Marije J. Zwart","email":"","orcid":"","institution":"Eindhoven University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Marije","middleName":"J.","lastName":"Zwart","suffix":""},{"id":375916103,"identity":"c57349ec-99ac-4370-8fa8-a74c73211b21","order_by":2,"name":"Arnoud W. Postema","email":"","orcid":"","institution":"Leiden University Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Arnoud","middleName":"W.","lastName":"Postema","suffix":""},{"id":375916104,"identity":"4a981f8f-226c-4970-a624-c2f4f16d529f","order_by":3,"name":"Daniel L. Kroonenberg","email":"","orcid":"","institution":"Amsterdam UMC location Vrije Universiteit Amsterdam","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"L.","lastName":"Kroonenberg","suffix":""},{"id":375916105,"identity":"745f3b42-179b-455c-9a28-47a11d38783d","order_by":4,"name":"Wim Zwart","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Wim","middleName":"","lastName":"Zwart","suffix":""},{"id":375916106,"identity":"ddf684a7-175d-4639-9133-124b8a1790a1","order_by":5,"name":"Harrie P. Beerlage","email":"","orcid":"","institution":"Amsterdam UMC, University of Amsterdam","correspondingAuthor":false,"prefix":"","firstName":"Harrie","middleName":"P.","lastName":"Beerlage","suffix":""},{"id":375916107,"identity":"5110ea37-dfe6-425f-919f-e063e018bc41","order_by":6,"name":"J. R. Oddens","email":"","orcid":"","institution":"Amsterdam UMC, University of Amsterdam","correspondingAuthor":false,"prefix":"","firstName":"J.","middleName":"R.","lastName":"Oddens","suffix":""},{"id":375916108,"identity":"ba7472c3-bbcd-4ba9-a899-509b7e70a3a2","order_by":7,"name":"Massimo Mischi","email":"","orcid":"","institution":"Eindhoven University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Massimo","middleName":"","lastName":"Mischi","suffix":""}],"badges":[],"createdAt":"2024-10-07 11:08:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5217620/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5217620/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12894-025-01736-4","type":"published","date":"2025-04-03T15:57:38+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":69950527,"identity":"1622aa48-764d-476b-a1fb-f6cbc4da13c3","added_by":"auto","created_at":"2024-11-27 02:12:43","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":627332,"visible":true,"origin":"","legend":"\u003cp\u003eRegistration process and accuracy calculation. 1. Transversal view of a CEUS and B-mode scan without (1a) and with (1b) segmentation of anatomical structures. Note the landmark (BPH nodule) clearly discernable on both TRUS and pathology (1b \u0026amp; 2b). 2. Pathology slides, including the parasagitally divided apical slice (2a) and the whole mount slides from apex to base (2b \u0026amp; 2c). 3. Registration with blue tension lines to visualise the losses. 4. Sparse, binary ground truth: white = benign, orange = malignant. 5. Accuracy calculation\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5217620/v1/2ff6788f7354fc7d28c03d75.png"},{"id":69950530,"identity":"a53e6e59-3bd6-49ab-b580-b87a8e11c5c4","added_by":"auto","created_at":"2024-11-27 02:12:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":235960,"visible":true,"origin":"","legend":"\u003cp\u003eImage acquisition sequence including acquisition time per ultrasound modality. Note that three separate 3D B-mode acquisitions are performed. These can be used for motion detection and compensation. B-mode = Brightness-mode, SWE = Shear Wave Elastography, IV = Intravenous, UCA = Ultrasound Contrast Agent.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5217620/v1/5dc93056671bb21877fab272.png"},{"id":69950531,"identity":"9293836a-742a-4c5b-86ab-3a3cadb3eb0a","added_by":"auto","created_at":"2024-11-27 02:12:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":430340,"visible":true,"origin":"","legend":"\u003cp\u003eprobe fixture. The probe fixture is attached to the DIN-rail and is used to fix the ultrasound probe in a position to ensure optimal and stable image acquisition.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5217620/v1/489a01910ae7d6d2c8fedfcc.png"},{"id":69950529,"identity":"c5e4d0c4-e365-417d-8ef9-1611ae299f48","added_by":"auto","created_at":"2024-11-27 02:12:44","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":54622,"visible":true,"origin":"","legend":"\u003cp\u003eRegistration flowchart. GT = Ground truth\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5217620/v1/37da198725833c04990eab89.png"},{"id":80082243,"identity":"42e1dba9-dd2b-4e6b-84da-02de721e03c7","added_by":"auto","created_at":"2025-04-07 16:07:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2165706,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5217620/v1/0d0e8dfa-44ed-4884-963f-7252b59ad06e.pdf"},{"id":69950528,"identity":"9aea0ab8-2ed1-478b-a010-bf0266786c28","added_by":"auto","created_at":"2024-11-27 02:12:44","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":459989,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Material\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupplementary materials 1: A visualization the ground truth for all 20 included patients. The 3D segmentations the ultrasound scans are visualized in red and green, red being the right prostate lobe and green the left. Within the prostate in yellow the border between the TZ and PZ is visualized, as well as the urethra and the spermatic ducts. The pathology slides are mapped onto the segmented ultrasound scans, the borders of which are shown as purple dots and blue lines.\u003c/p\u003e","description":"","filename":"Supplementarymaterials1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5217620/v1/8f92b938a987dfa53f3e20b8.pdf"}],"financialInterests":"Competing interest reported. Conflict of Interest:\n\nHB is chair of the clinical board for AA.\nAP and MM are scientific advisors for AA for which they receive compensation.\nMZ and WZ are employees of AA.","formattedTitle":"Development and Validation of a Framework for Registration of Whole-Mount Radical Prostatectomy Histopathology with Three-dimensional Transrectal Ultrasound","fulltext":[{"header":"Introduction","content":"\u003cp\u003eImaging plays an increasingly important role in prostate cancer (PCa) diagnostics. Prostate magnetic resonance imaging (MRI) prior to prostate biopsy is currently the standard of care for patients with a clinical suspicion for PCa (1). While MRI has greatly improved the diagnostic pathway for PCa, it still has certain limitations related to diagnostic accuracy, availability, cost, and interobserver agreement (2, 3).\u003c/p\u003e \u003cp\u003eAs a result, various methods to improve prostate imaging are being investigated. These methods are related to advancements in MRI performance and efficiency, as well as the exploration of alternative imaging modalities such as transrectal ultrasound (TRUS) and prostate-specific membrane antigen (PSMA) positron emission tomography (PET)-based modalities (4\u0026ndash;6). An area of particular interest is the integration of artificial intelligence (AI) into the diagnostic process, which shows promising potential in improving diagnostic accuracy, reducing interobserver variability, and addressing time-intensity concerns (7\u0026ndash;9)\u003c/p\u003e \u003cp\u003eRegarding TRUS as an alternative modality, the multiparametric approach shows promise. This approach combines information from different ultrasound modalities, including Brightness-mode (B-mode), contrast-enhanced ultrasound (CEUS), and shear wave elastography (SWE). Machine learning (ML) techniques have emerged as a valuable tool for integrating and extracting information from these ultrasound modalities (5, 8, 10).\u003c/p\u003e \u003cp\u003eIn recent developments, the availability of 3D TRUS has provided further support to these advancements. The capability of imaging the entire prostate gland in a single scan offers evident advantages for the clinical workflow (11\u0026ndash;13). (3D) TRUS-based imaging may be advantageous over MRI due to its cost-effectiveness and wider availability, although its value in clinical practice for PCa diagnosis is not yet established (5).\u003c/p\u003e \u003cp\u003eOngoing research aims to develop AI methods to detect PCa on different imaging modalities (11). For development and validation of such methods, a reliable ground truth is required (14). The best way to obtain this ground truth is through correlation of pre-surgical in-vivo imaging with pathology slides obtained from radical prostatectomy specimens (RPS) (15). The pathology slides contain information on the localization and characterization of cancerous tissue, which can then be mapped onto the corresponding images recorded in-vivo. However, correlation of pathology to imaging (i.e., registration) poses several challenges.\u003c/p\u003e \u003cp\u003eFollowing surgery, the RPS is fixated in formalin and sliced into slices of approximately equal size, which are then used to create pathology slides for evaluation by a pathologist. Due to variation in slice thickness, differences between pathology and imaging slice orientation, and RPS deformation, pathology and imaging slices cannot simply be overlaid for correlation (16). For MRI, multiple registration approaches using open-source ML algorithms have been proposed that reach fairly high levels of accuracy (15, 17). These approaches cannot be applied to TRUS due to a different slice orientation and prostate deformation due to the pressure applied by the ultrasound probe on the prostate (16).\u003c/p\u003e \u003cp\u003ePrevious work has shown promising results using affine and elastic, surface-based registration for in-vivo registration of 2D TRUS with RPS pathology, with a mean target registration error (TRE) of 2.1 mm (18). However, the validation was limited to a small cohort of seven patients. In contrast, the current study employs a larger dataset and uses 3D TRUS, providing a more robust validation. This paper presents a new framework for the registration of 3D TRUS with 3D RPS pathology and reports on its in-vivo validation. Although the framework is primarily developed for TRUS, it proposes several innovative techniques that can potentially facilitate accurate registration of pathology across different imaging modalities.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eThe cohort consisted of patients planned for radical prostatectomy (RP) because of biopsy-proven PCa, prospectively included in a multicentre study. The trial protocol for this study has been published (11). The framework for TRUS-pathology registration can be separated in four steps 1) TRUS acquisition and segmentation 2) RPS pathology handling and 3D reconstruction, 3) TRUS-pathology registration and 4) calculation of ground truth (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e1. \u003cstrong\u003eThree-dimensional transrectal prostate ultrasound acquisition and segmentation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll patients underwent 3D multiparametric transrectal ultrasound (3D mpUS) prior to RP using a LOGIQ\u003csup\u003eTM\u003c/sup\u003e E10 ultrasound machine with a RIC 5-9D 3D endocavitary probe (GE Healthcare, Chicago, IL, USA). A motorized mechanism inside of the probe supports the automatic acquisition of a 120\u0026deg; volumetric sweep. 3D mpUS acquisition consisted of 3D B-mode, 3D shear-wave elastography, and 4D contrast enhanced imaging and is described in detail in the published trial protocol (Fig 2) (11). To ensure stable image acquisition, a fixture device was developed to fixate the ultrasound probe after satisfactory positioning (Fig 3). \u003c/p\u003e\n\u003cp\u003eImage processing consisted of manual segmentation of anatomical structures, using a segmentation tool developed specifically for this study by Angiogenesis Analytics (AA, \u0026lsquo;s-Hertogenbosch, the Netherlands). Segmentations were performed by AJ or AP, with respectively 4 and 8 years of experience with TRUS. Standard segmentation consisted of the prostate border, the border between peripheral zone (PZ) and transitional zone (TZ), and the urethra. Additional segmentation of other anatomical structures (utriculus and ejaculatory ducts) and landmarks (e.g., calcifications and cysts) were only performed when clearly visible on both ultrasound and pathology. \u003c/p\u003e\n\u003cp\u003e2. \u003cstrong\u003eRadical prostatectomy pathology\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e2a. Radical prostatectomy specimen handling\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eRPS handling was performed according to a study specific protocol (11). To facilitate accurate 3D pathology reconstruction and subsequent registration to TRUS, certain steps were included in the protocol. Prior to fixation in formalin, four intravenous cannulas were inserted in four different quadrants of the RPS from apex to base. After at least 24h of fixation in formalin, the cannulas were removed, the RPS was cut into 4-mm thick slices from apex to base using the TruSlice specimen cut up system (CellPath ltd, Newtown, UK) and a photo of the gross pathology slices was captured. The most apical and basal slice were further divided parasagittally in 4-mm strips. Pathology slides were cut at 4-\u0026micro;m thickness from the gross pathology slices using a microtome and scanned on high resolution (40x enlargement, 20x objective, 2.1 camera lens) with a Pannoramic 1000 Digital Slide Scanner (3DHISTECH, H-1141 Budapest, Hungary). Digital pathology slices were annotated according to study protocol by a pathologist using a web-based pathology annotation tool (Slidescore, Amsterdam, the Netherlands)(11). Apart from cancerous areas, the annotation protocol also included the same anatomical structures and landmarks segmented on TRUS. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003e2b. Three-dimensional pathology reconstruction\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAfter annotation, pathology slides were overlaid on the gross image of prostate slices. The gross prostate slices with overlay were then stacked to create a 3D pathology model. The needle holes resulting from the intravenous cannulas were used to prevent rotational and translational errors between the slices. Shrinkage of histopathology slides could be compensated for up to 10%. The apical pathology slices were sequenced from left to right, checked for anterior and posterior orientation, and fitted on the most apical gross prostate slice. The parasagittal pathology slides from the most basal (or bladder neck) slice were not included in the reconstruction. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Registration of pathology to ultrasound images\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe registration is defined by both a rigid and a non-rigid transform. By applying first, the rigid, and then the non-rigid transform, the pathology slides are mapped onto the TRUS images, minimising the distance between the visible anatomical structures in both domains. The registration procedure compensates for in-vivo deformation of the prostate caused by pressure from the ultrasound probe, as well as ex-vivo deformation resulting from the excision of the prostate and subsequent fixation in formalin (19). The full registration algorithm was implemented and visualised in a registration tool (developed by AA), with the individual steps indicated in Fig 4.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3a. Compensation for Probe Deformation\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eDuring TRUS acquisition, the prostate is deformed in-vivo due to pressure applied by the probe. As this is not present in the pathology slices, compensation of this deformation is required for accurate registration. This is done by modelling the deformation effect of the probe on homogeneous tissue based on a 3D scan of the probe. The inverse of this deformation is then applied to the 3D ultrasound segmentations. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3b. Rigid Registration\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe first step of the registration is to define a rigid transform, described by a translation, a rotation and three scaling factors. The combination of translation and rotation parameters positions the full reconstructed prostate to the ultrasound scan domain. Two of the scaling factors effectively act as gross and apex slice thicknesses, and the third is applied to all slices in their respective surface planes to account for the aforementioned shrinking factors. \u003c/p\u003e\n\u003cp\u003eThe first step in defining this rigid transform, is to align the centre of the ultrasound-scan boundary segmentation with the centre of the boundary annotations from the pathology slices as well as aligning the left-right, anterior-posterior, and apex-base directions in both pathology and scan domain. Subsequently, the parameters are optimized by an iterative closest point (ICP) algorithm. For each visible anatomical structure, a set of point-pairs is defined by pairing all points of the annotations of the respective structure in the pathology domain with the closest point on the segmentations in the scan domain. A loss term for each anatomical structure is then calculated as the root mean square error (RMSE) of the distances between the points in each pair. The total loss is now defined as the weighted sum of the individual loss terms of each set. The transform with minimum total loss is found through gradient-descent and applied to the points in the pathology domain. Subsequently, a new set of closest points is found, and the process repeated until the minimised total loss has converged. The blue \u0026lsquo;tension\u0026rsquo; lines in the registration section of Fig. 1 present a visual representation of the point-pairs. \u003c/p\u003e\n\u003cp\u003eAs mentioned, the parasagittal pathology slides from the most basal gross slice are not included in the reconstruction. This leads to a significant degree of freedom in the placement of the gross slices over the apex-base direction. As this degree of freedom is poorly reflected in the previously described loss terms, an additional loss term is included to reflect the difference between a desired base height and the effective base height resulting from the current rigid transform parameters. \u003c/p\u003e\n\u003cp\u003eThe registration tool provides for means for an operator to monitor the registration process and manually adjust some of the key parameters. These adjustable parameters include weight factors of the described loss terms, a desired base height and adjustment for shear. The weight factors of the anatomical structures are adjusted to aim for a homogeneous distribution of the loss over each structure. The desired base height is adjusted based on visual inspection of tension lines, aiming to settle at a gross slice thickness of to account for the shrinking mentioned earlier. Any indication of shear in the stack of pathology slices is also manually adjusted for. The optimization iterations are terminated once the total loss has converged, and the operator recognises no need for further adjustments. To account for the operational variation, this step was reviewed by a 2nd operator and discussed where necessary to reach consensus. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3c. Non-Rigid Registration\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe second step of the registration is to define a non-rigid transform which is applied after the rigid transform. The non-rigid transform compensates for local deformation, assuming homogeneous elasticity of the tissue. Following the concept of the Finite Element Method (FEM), the scan space is subdivided into a regular grid with a size of 2mm. The deformation (i.e. variation in distance) between adjacent grid points is calculated, resulting in one overall additional loss term for elasticity. Finding the deformation that minimizes the elasticity loss term is like mimicking the behaviour of elastic tissue that offers some degree of resistance against deformation. The non-rigid registration phase has now become similar to the rigid registration phase, with the elasticity loss that is added the overall weight with an appropriate weight as the key difference. \u003c/p\u003e\n\u003cp\u003eIterations of the optimization algorithm are carried out until either a) the difference in loss between one iteration and the next was smaller than a pre-defined value or b) a maximum number of iterations was reached. Note that no input of the operator is required in this step. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. Calculation of ground truth\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eApplying both the rigid and non-rigid transforms, the pathology annotations are mapped onto the 3D ultrasound scan space which was compensated for probe deformation. Therefore, the last step consists of applying the probe deformation to the annotations. Voxels lying inside the registered slides can now be assigned as either malignant or benign, based on the pathology annotations. Note that this does not provide any information on the voxels lying in between the pathology slides. Interpolation techniques are applied to the voxels lying in between the pathology slides to assign each of these with a probability of their being malignant. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatient selection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients who exhibited a clearly distinguishable landmark in both the pathology and ultrasound scan domains were selected for the assessment of registration accuracy. The determination of the validity of these landmarks was made through consensus between two independent observers (AJ \u0026amp; AP). \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRegistration accuracy\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn order to calculate the target-registration error (TRE), the registration was performed with deliberate exclusion of the landmark as an input to the optimization algorithm, thereby mitigating its potential influence on the registration outcome. The landmarks could then be used to calculate the TRE between pathology and ultrasound. \u003c/p\u003e\n\u003cp\u003eFollowing registration, each landmark annotation on the pathology slides results in a group of registered landmark voxels in ultrasound scan domain, of which the outer points were extracted. Subsequently, an ICP algorithm, similar to the one used in the registration process, was employed to obtain a translation that minimises the difference between the transformed outer voxel points and points on the landmark contour(s) obtained from the scan. To establish point pairs, the untransformed outer voxel point was paired with the point on the landmark contour lying closest to the translated outer voxel point. The TRE was then defined as the mean of the Euclidean distances between these point pairs. The magnitude of the TRE in the apex-base (AB), left-right (LR) and posterior-anterior (PA) axes only was also evaluated by projecting the distances onto each axis. \u003c/p\u003e\n\u003cp\u003eThe TRE was determined for each patient, and the overall registration accuracy was subsequently expressed as the mean of these TREs.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe total cohort consisted of 126 eligible patients, of which 20 exhibited a clearly discernable landmark and were included in the registration accuracy analysis.\u003c/p\u003e \u003cp\u003eThe overall registration accuracy (mean TRE) was 3.5 mm (range 0.4 mm \u0026ndash; 5.4 mm), with a TRE in the AB, LR and PA directions of 2.5, 1.1 and 1.4 respectively. Table\u0026nbsp;1 gives an overview of the per patient TRE and types of landmarks. A visualization of all ultrasounds, pathology slides and registrations, including their landmarks is provided in supplementary materials 1.\u003c/p\u003e \u003cp\u003e\u003cstrong\u003eTable 1\u003c/strong\u003e\u0026nbsp; Per patient TRE. \u0026nbsp;TRE = target registration error, AB = apex-base, LR = left-right, PA = posterior-anterior, BPH = benign prostatic hyperplasia.\u003c/p\u003e\u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProstate volume (mL)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLandmark\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTRE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTRE (AB)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTRE (LR)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eTRE (PA)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e133.06\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e28.39\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e33.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e59.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBPH nodule\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e88.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e35.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e80.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e45.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e45.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e90.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e76.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e28.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e62.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e100.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBPH Nodule\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e47.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e28.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e25.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e47.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e28.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e27.29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e57.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedian\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e46.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe registration of RPS pathology to prostate imaging poses a considerable challenge, particularly in the context of TRUS. The current study presents a comprehensive framework for accurate registration of RPS to 3D TRUS, with a mean TRE of 3.5 mm based on 20 patients with clearly discernable landmarks. While there is no consensus on the maximum acceptable registration error, clinically significant PCa lesions are generally considered to have a volume of \u0026ge;\u0026thinsp;0.5 cm\u003csup\u003e3\u003c/sup\u003e, which corresponds to a diameter of 10 mm, assuming to lesion is spherical (16). Therefore, we should aim to have a registration error below 10 mm.\u003c/p\u003e \u003cp\u003eThis study makes significant advancements compared to existing literature in several areas.\u003c/p\u003e \u003cp\u003ePrior research has demonstrated the feasibility of registering pathology to ultrasound with acceptable levels of accuracy (18). Our study builds upon this foundation by employing a larger dataset and a more comprehensive methodology. Notably, our study represents the first use of 3D acquired TRUS imaging, as opposed to relying on a 3D reconstruction from handheld 2D transversal sweeps. To ensure precise 3D RPS reconstruction, we implemented needle placement prior to RPS fixation, effectively preventing transitional and rotational errors. Furthermore, to aid the registration, we included compensation for the probe deformation. Although several assumptions and simplifications had to be made to model the deformation, it has shown to be a valuable addition to the registration process. Furthermore, while the use of anatomical structures for improving registration accuracy has generally been limited to the prostate border and urethra (15, 16, 18). Based on our experience the addition of the border between PZ and TZ is feasible for most patients and is highly useful during registration. Although the visibility of ejaculatory ducts can vary (7 out of 20 cases in our dataset), when available, they offer valuable information during registration. When using ultrasound as the imaging modality, it is important to note that these ducts are generally only visible on CEUS and not on conventional B-mode. Additionally, landmarks had to be excluded from the registration to be able to calculate accuracy, but when included, they can provide information on slice and base thickness, and thus have a potential to improve accuracy. In our set of 20 patients, we had 3 patients with 2 landmarks available. For these patients, we repeated the registration but this time including the landmark that had not been used for validation. The new results and their differences compared to the old results are shown in table 2.\u003c/p\u003e \u003cp\u003e\u003cstrong\u003eTable 2\u003c/strong\u003e This table illustrates the effect of landmarks on the TRE. The values within the parentheses indicate change in TRE resulting from the integration of a landmark into the registration procedure. \u0026nbsp;TRE = target registration error, AB = apex-base, LR = left-right, PA = posterior-anterior, BPH = benign prostatic hyperplasia.\u003c/p\u003e\u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabb\" border=\"1\"\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTRE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTRE (AB)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTRE (LR)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTRE (PA)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e \u003cp\u003e4.9 (-0.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e3.6 (-0.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.6 (+\u0026thinsp;0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.6 (+\u0026thinsp;0.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e \u003cp\u003e2.7 (-1.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e2.3 (-1.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.6 (-0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.4 (+\u0026thinsp;0.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e \u003cp\u003e2.3 (-2.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e0.9 (-2.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.7 (+\u0026thinsp;0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.2 (-2.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThis is a very small set of patients and therefore caution should be taken with any conclusions. However, all patients showed a significant overall improvement by inclusion of a landmark in the registration, especially in the apex-base direction.\u003c/p\u003e \u003cp\u003eLastly, our study demonstrates the feasibility of including parasagittally cut apical slides in the pathology reconstruction, which has not been previously described due to the complexities involved in reconstruction (16). These slides offer valuable information, as they often contain PCa lesions (20). In fact, in our dataset, 75% of patients (15 out of 20) exhibited some form of prostate cancer in the apical pathology slides, underscoring the importance of including these slides in a ground truth dataset.\u003c/p\u003e \u003cp\u003eTo optimize registration accuracy, several actions were taken during the handling of prostate specimens. A commercially available cutting device was employed to ensure uniform thickness of pathology slices, and needle placement prior to fixation facilitated proper stacking of the slides after slicing.\u003c/p\u003e \u003cp\u003eThe biggest uncertainty in the registration is in the apex-base axis, as we have the least information in that direction. To add further complexity, since the length of the prostate from apex to base may not be evenly divisible by 4, and the apex is cut from apex to base, the thickness of the most basal slide often differs from the other slices. Although the inclusion of the apical slides aids in appropriately positioning the whole-mount slices from apex to base towards the apex, the most basal slides (i.e., bladder neck) were not included in the reconstruction process. The decision to exclude these slides stems from several factors. Firstly, the basal slice is typically larger than the apical slice and is cut thicker to facilitate the evaluation of extracapsular extension. Consequently, pathology slides originating from the basal slice are numerous and often need to be further divided to fit into pathology cassettes.\u003c/p\u003e \u003cp\u003eMoreover, the identification of the prostate border becomes more challenging in the vicinity of the urethra and seminal vesicles, further complicating the reconstruction of the basal slide. Additionally, it is worth mentioning that the incidence of prostate cancer in the base is generally low, further justifying the exclusion of the basal slide from the reconstruction process (21).\u003c/p\u003e \u003cp\u003e(22). However, this leaves a significant range of possible values for the slice and base thicknesses, with the final chosen value dependent on the operator. This is reflected in our TRE calculations, which show that the TRE in the apex-base direction is twice as large as in the left-right and posterior-anterior directions.\u003c/p\u003e \u003cp\u003eSeveral limitations should be acknowledged in this study. Firstly, the manual segmentation, annotation and registration involved are time intensive, costing approximately an hour per prostate. Efforts to automate segmentation of ultrasound images through deep learning algorithms have shown promising results and are expected to reduce the required manual labor. Moreover, the manual nature of these procedures introduces the possibility of interobserver variability. Certain ultrasound segmentations, such as those for the urethra and zones, may not always be clearly visible, particularly in cases involving larger prostates. Although consensus meetings were held among operators to address uncertainties in specific segmentations, it is important to acknowledge that this process may have influenced the registration accuracy to some extent. A suggested improvement to the registration algorithm would be to provide it with a weighting factor for each segmentation point based on the quality of the scan at that location. This would allow the registration to put more focus on more clearly visible \u0026ndash; and therefore more certain - areas of the US segmentations.\u003c/p\u003e \u003cp\u003eLastly, the registration method employed in this study requires custom-made software for execution, which may present challenges when implementing it elsewhere. However, the authors can be contacted for assistance in research projects, and they are willing to share data upon reasonable request, facilitating further exploration and collaboration in the field.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe study presents a comprehensive framework for registering RPS pathology to 3D TRUS, achieving accurate results based on discernable landmarks. Precise registration is crucial for generating reliable datasets, essential in developing and validating new diagnostic tools and for training AI-assisted imaging modalities.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eConflict of Interest:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHB is chair of the clinical board for AA.\u003c/p\u003e\n\u003cp\u003eAP and MM are scientific advisors for AA for which they receive compensation.\u003c/p\u003e\n\u003cp\u003eMZ and WZ are employees of AA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval and Consent to Participate\u003c/strong\u003e: this study was approved by an accredited medical research ethics committee (MEC AMC) under reference number 2020_268#B202178. All study participants signed an informed consent form that includes the consent for use of their data for publication. The study was performed in accordance with the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAJ: Conceptualization, Methodology, Acquisition, Writing – Original Draft, Visualization, Project administration. MZ: conceptualization, Methodology, Writing – Original Draft, Visualization, Project administration, interpretation of data, statistical analysis. DK \u0026amp; AP: Conceptualization, Methodology, Acquisition, Writing – Review \u0026amp; Editing. WZ: Methodology, interpretation of data, statistical analysis. JO: Writing – Review \u0026amp; Editing, Supervision, Funding acquisition. HB: Writing – Review \u0026amp; Editing, Supervision, Funding acquisition. MM: Conceptualization, Methodology, Writing – Review \u0026amp; Editing, Supervision, Project administration.\u0026nbsp;All authors read and approved the final paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was co-funded by the European Union and Angiogenesis Analytics.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e: not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eMottet N, van den Bergh RCN, Briers E, Van den Broeck T, Cumberbatch MG, De Santis M, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol. 2021;79(2):243-62.\u003c/li\u003e\n\u003cli\u003eDrost FH, Osses D, Nieboer D, Bangma CH, Steyerberg EW, Roobol MJ, et al. Prostate Magnetic Resonance Imaging, with or Without Magnetic Resonance Imaging-targeted Biopsy, and Systematic Biopsy for Detecting Prostate Cancer: A Cochrane Systematic Review and Meta-analysis. Eur Urol. 2020;77(1):78-94.\u003c/li\u003e\n\u003cli\u003eBrembilla G, Dell\u0026apos;Oglio P, Stabile A, Damascelli A, Brunetti L, Ravelli S, et al. Interreader variability in prostate MRI reporting using Prostate Imaging Reporting and Data System version 2.1. Eur Radiol. 2020;30(6):3383-92.\u003c/li\u003e\n\u003cli\u003eLisney AR, Leitsmann C, Strauss A, Meller B, Bucerius JA, Sahlmann CO. The Role of PSMA PET/CT in the Primary Diagnosis and Follow-Up of Prostate Cancer-A Practical Clinical Review. Cancers (Basel). 2022;14(15).\u003c/li\u003e\n\u003cli\u003eMannaerts CK, Engelbrecht MRW, Postema AW, van Kollenburg RAA, Hoeks CMA, Savci-Heijink CD, et al. Detection of clinically significant prostate cancer in biopsy-naive men: direct comparison of systematic biopsy, multiparametric MRI- and contrast-ultrasound-dispersion imaging-targeted biopsy. BJU Int. 2020;126(4):481-93.\u003c/li\u003e\n\u003cli\u003eTurkbey B, Haider MA. Deep learning-based artificial intelligence applications in prostate MRI: brief summary. Br J Radiol. 2022;95(1131):20210563.\u003c/li\u003e\n\u003cli\u003ede Rooij M, van Poppel H, Barentsz JO. Risk Stratification and Artificial Intelligence in Early Magnetic Resonance Imaging-based Detection of Prostate Cancer. Eur Urol Focus. 2021.\u003c/li\u003e\n\u003cli\u003eWildeboer RR, Mannaerts CK, van Sloun RJG, Budaus L, Tilki D, Wijkstra H, et al. Automated multiparametric localization of prostate cancer based on B-mode, shear-wave elastography, and contrast-enhanced ultrasound radiomics. Eur Radiol. 2020;30(2):806-15.\u003c/li\u003e\n\u003cli\u003eYi Z, Hu S, Lin X, Zou Q, Zou M, Zhang Z, et al. Machine learning-based prediction of invisible intraprostatic prostate cancer lesions on (68) Ga-PSMA-11 PET/CT in patients with primary prostate cancer. Eur J Nucl Med Mol Imaging. 2022;49(5):1523-34.\u003c/li\u003e\n\u003cli\u003eSuarez-Ibarrola R, Sigle A, Eklund M, Eberli D, Miernik A, Benndorf M, et al. Artificial Intelligence in Magnetic Resonance Imaging-based Prostate Cancer Diagnosis: Where Do We Stand in 2021? Eur Urol Focus. 2021.\u003c/li\u003e\n\u003cli\u003eJager A, Postema AW, Mischi M, Wijkstra H, Beerlage HP, Oddens JR. Clinical Trial Protocol: Developing an Image Classification Algorithm for Prostate Cancer Diagnosis on Three-dimensional Multiparametric Transrectal Ultrasound. Eur Urol Open Sci. 2023;49:32-43.\u003c/li\u003e\n\u003cli\u003eWildeboer RR, van Sloun RJG, Huang P, Wijkstra H, Mischi M. 3-D Multi-parametric Contrast-Enhanced Ultrasound for the Prediction of Prostate Cancer. Ultrasound Med Biol. 2019;45(10):2713-24.\u003c/li\u003e\n\u003cli\u003eWildeboer RR, Van Sloun RJG, Schalk SG, Mannaerts CK, Van Der Linden JC, Huang P, et al. Convective-Dispersion Modeling in 3D Contrast-Ultrasound Imaging for the Localization of Prostate Cancer. IEEE Trans Med Imaging. 2018;37(12):2593-602.\u003c/li\u003e\n\u003cli\u003eWillemink MJ, Koszek WA, Hardell C, Wu J, Fleischmann D, Harvey H, et al. Preparing Medical Imaging Data for Machine Learning. Radiology. 2020;295(1):4-15.\u003c/li\u003e\n\u003cli\u003eRusu M, Shao W, Kunder CA, Wang JB, Soerensen SJC, Teslovich NC, et al. Registration of presurgical MRI and histopathology images from radical prostatectomy via RAPSODI. Med Phys. 2020;47(9):4177-88.\u003c/li\u003e\n\u003cli\u003eWildeboer RR, van Sloun RJG, Postema AW, Mannaerts CK, Gayet M, Beerlage HP, et al. Accurate validation of ultrasound imaging of prostate cancer: a review of challenges in registration of imaging and histopathology. J Ultrasound. 2018;21(3):197-207.\u003c/li\u003e\n\u003cli\u003eShao W, Banh L, Kunder CA, Fan RE, Soerensen SJC, Wang JB, et al. ProsRegNet: A deep learning framework for registration of MRI and histopathology images of the prostate. Med Image Anal. 2021;68:101919.\u003c/li\u003e\n\u003cli\u003eSchalk SG, Postema A, Saidov TA, Demi L, Smeenge M, de la Rosette JJ, et al. 3D surface-based registration of ultrasound and histology in prostate cancer imaging. Comput Med Imaging Graph. 2016;47:29-39.\u003c/li\u003e\n\u003cli\u003eJonmarker S, Valdman A, Lindberg A, Hellstrom M, Egevad L. Tissue shrinkage after fixation with formalin injection of prostatectomy specimens. Virchows Arch. 2006;449(3):297-301.\u003c/li\u003e\n\u003cli\u003eVeerman H, Boellaard TN, van Leeuwen PJ, Vis AN, Bekers E, Hoeks C, et al. The detection rate of apical tumour involvement on preoperative MRI and its impact on clinical outcomes in patients with localized prostate cancer. J Robot Surg. 2022;16(5):1047-56.\u003c/li\u003e\n\u003cli\u003eVargas SO, Jiroutek M, D\u0026apos;Amico AV, Renshaw AA. Distribution of carcinoma in radical prostatectomy specimens in the era of serum prostate-specific antigen testing. Implications for delivery of localized therapy. Am J Clin Pathol. 1999;112(3):373-6.\u003c/li\u003e\n\u003cli\u003eSakamoto Y, Fukaya K, Haraoka M, Kitamura K, Toyonaga Y, Tanaka M, et al. Analysis of prostate cancer localization toward improved diagnostic accuracy of transperineal prostate biopsy. Prostate Int. 2014;2(3):114-20.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-urology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"buro","sideBox":"Learn more about [BMC Urology](http://bmcurol.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/buro/default.aspx","title":"BMC Urology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5217620/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5217620/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eArtificial intelligence (AI) has the potential to improve diagnostic imaging on multiple levels. To develop and validate these AI-assisted modalities a reliable dataset is of utmost importance. The registration of imaging to pathology is an essential step in creating such a dataset. This study presents a comprehensive framework for the registration of 3D transrectal ultrasound (TRUS) to radical prostatectomy specimen (RPS) pathology.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study enrolled patients who underwent 3D TRUS and were scheduled for radical prostatectomy. A four-step process for registering RPS to TRUS was used: image segmentation, 3D reconstruction of RPS pathology, registration and ground truth calculation. Accuracy was assessed using a target-registration error (TRE) based on landmarks visible on both TRUS and pathology.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e20 Sets of 3D TRUS and RPS pathology were included for analyses. The mean TRE was 3.5 mm, ranging from 0.4 to 5.4 mm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe framework proposed in this study accomplishes precise registration between prostate pathology and imaging. The methodologies employed hold the potential for broader application across diverse imaging modalities and other target organs.\u003c/p\u003e","manuscriptTitle":"Development and Validation of a Framework for Registration of Whole-Mount Radical Prostatectomy Histopathology with Three-dimensional Transrectal Ultrasound","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-27 02:12:39","doi":"10.21203/rs.3.rs-5217620/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorAssigned","content":"","date":"2024-11-07T07:06:15+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-11-01T02:04:41+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Urology","date":"2024-10-07T11:00:07+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-urology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"buro","sideBox":"Learn more about [BMC Urology](http://bmcurol.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/buro/default.aspx","title":"BMC Urology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c08f0e59-3c04-4d0b-9db3-6e9e8313e2b2","owner":[],"postedDate":"November 27th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-04-07T16:04:21+00:00","versionOfRecord":{"articleIdentity":"rs-5217620","link":"https://doi.org/10.1186/s12894-025-01736-4","journal":{"identity":"bmc-urology","isVorOnly":false,"title":"BMC Urology"},"publishedOn":"2025-04-03 15:57:38","publishedOnDateReadable":"April 3rd, 2025"},"versionCreatedAt":"2024-11-27 02:12:39","video":"","vorDoi":"10.1186/s12894-025-01736-4","vorDoiUrl":"https://doi.org/10.1186/s12894-025-01736-4","workflowStages":[]},"version":"v1","identity":"rs-5217620","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5217620","identity":"rs-5217620","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}