Digital twins for breast cancer surgery

Digital twins for breast cancer surgery | 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 Article Digital twins for breast cancer surgery Rafaela Timóteo, Alexandre Laborde, Bruno Vaz, Nuno Loução, Yasna Forghani, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8907845/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Breast-conserving surgery followed by adjuvant radiotherapy is the standard treatment for breast cancer, yet inaccurate tumor localization may lead to positive margins and re-excision. Current localization techniques are invasive and often imprecise. Digital twins (DT), patient-specific digital replicas, can offer a non-invasive approach for tumor localization. In this proof-of-concept study, breast DTs were created from supine magnetic resonance imaging and visualized through an extended reality (XR) headset in four breast conservative surgeries. DTs were automatically registered to the patient in real time using a 3D sensing system. The accuracy of the alignment deviations between the nipple and its position in the DT ranged 0-5 mm. The surgeon reported that the tumor XR visualization was consistent with intraoperative findings. Setup and marking times were clinically acceptable, and preliminary usability assessments indicated high perceived usability and low workload. These results support the feasibility of integrating DTs into breast cancer surgical workflows. Biological sciences/Cancer Health sciences/Health care Health sciences/Medical research Health sciences/Oncology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Breast cancer (BC) is the second most common cancer worldwide, with approximately 2.3 million new cases reported annually [1,2]. In Western countries, the widespread implementation of screening programs has increased the rate of early diagnosis. As a result, breast-conserving surgery (BCS) followed by adjuvant radiotherapy has become the gold standard for local treatment, accounting for 60–80% of newly diagnosed cases in 2022 [3]. Early detection often involves small, non-palpable lesions, which complicates surgical management. Successful BCS relies on precise tumor localization and complete excision with clear margins to minimize the risk of re-excision [4,5]. Surgical planning, decision-making, and outcomes rely on imaging modalities, with the most common being breast magnetic resonance imaging (MRI), acquired in the prone position. Due to the deformable nature of the breast tissue, and position change between imaging and surgery, planning is particularly challenging [6,7]. Localization of non-palpable lesions requires invasive techniques, including wire-guided, carbon tattooing, biopsy clips, radioactive seed localization, radio-occult lesion localization, and magnetic seeds [4]. In recent years, non-wired, non-ionizing devices have emerged [8], and supine MRI has been proposed as a solution to better approximate the patient’s surgical position [9,10]. Despite these advances, conventional tumor localization approaches remain percutaneous, often preoperative, uncomfortable for patients, and lacking optimal precision. Digital twins (DTs) are emerging in healthcare as patient-specific dynamic digital replicas, integrating both anatomical and physiological information, as a powerful tool for screening, diagnosis, treatment, decision-making, simulations, and outcome predictions [11, 12]. Generated from medical imaging, such as MRI, DTs can be aligned with the anatomy in the operating room (OR) and visualized through extended reality (XR) head-mounted displays (HMDs) to guide interventions [13,14]. Integrating DTs and XR in surgery offers digital, non-invasive guidance with benefits for both patient satisfaction and surgical performance [15]. DTs can be classified by the data they contain and how frequently they are updated. Most commercial solutions are centered on static twins, which do not change over time [12]. By incorporating finite element algorithms, functional twins can simulate the tissue behavior under specific conditions [12]. Shadow (self-adaptive) twins integrate real-time sensors or imaging data to account for tissue deformation and respiratory motion, optimizing surgical workflows [2,12]. Intelligent twins extend this further by providing continuous feedback and real-time risk prediction, simulations, and surgical planning, often leveraging biometric devices, implantable sensors, and machine learning [11,12]. Despite growing research, intraoperative use of DTs in XR remains largely experimental, with most work to date focused on orthopedics or neurosurgery [12-16]. In breast cancer surgical treatment, studies of such character remain scarce [17-19]. In this study, we present a DT-based XR navigation system for BCS tumor localization in the OR and perform a proof-of-concept pilot study to validate its performance. Breast DTs were created from segmented supine breast MRI images, aligned to the patient’s torso using color and depth sensors, and displayed to the surgeon through an XR headset during BCS (Figure 1). This proof-of-concept study provides initial evidence of the feasibility of an XR-based, non-invasive tumor localization in BCS. Results Cohort Four (4) breast cancer patients scheduled for breast surgery were recruited via a signed informed consent form at the Breast Unit at Champalimaud Foundation, between July and November 2025. The median patient age was 48 years (range 44-51). In addition to the standard-of-care breast MRI acquisition in prone, patients underwent breast MRI in supine (Figure 2a). Breast supine MRIs were segmented into three tissues: fat, fibroglandular tissue (FGT), and tumor (Figure 2b). One patient had the tumor in the left breast, and the other three in the right breast. Tumor volume and oriented bounding box (OBB) diameters were calculated from the segmentations (Table 1). One patient had a large tumor (maximum diameter of 74.2 mm) and underwent mastectomy followed by breast reconstructive surgery. Another patient had a very small satellite lesion near the primary tumor (Table 1). Breast static DTs were created from the segmented MRI images after post-processing (Figure 2c). Table 1. Patients’ lesion description after segmentation. Patient #2 had a large lesion and underwent a mastectomy. Patient #3 had a small satellite lesion near the primary tumor. Patient #3 had the tumor on the left breast, while the others had the tumor on the right breast. Patient # Breast with tumor Tumor volume (cm³) OBB diameter - WxHxL (mm) 1 Right 0.23 5.5 x 8.1 x 11.7 2 Right 12.74 35.9 x 50.8 x 74.2 3 Left 0.25 6.5 x 10.9 x 13.4 0.02 4.0 x 4.5 x 5.3 4 Right 1.66 14.4 x 15.2 x 21.3 Tumor marking in the operating room The system was set up after patient anesthesia, with the setup time taking 218 seconds (range 72-190). During setup, the surgeon was asked to put on the Magic Leap 2 (ML) headset and complete a standard eye-calibration process. Then, the surgeon entered the XR application and the patient’s DT was displayed and aligned within seconds (Figure 3). The visualization mode in XR contained the tumor in red, FGT in blue, nipples in yellow, a 1cm grid along the patient’s skin in white, and the tumor projection on the skin in pink. Tumor marking took on average 49 seconds (range 32-72). Across patients, the digital nipple deviation ranged from 0 mm to 5 mm, indicating a very good alignment between the breast DT and the patient’s anatomy in the OR (Figure 3). Tumor skin projection deviations relative to carbon tattoo markings ranged from 0 mm to large deviations (> 30 mm) (Table 2). In two cases where large deviations were detected, the intraoperative tumor position was more vertically aligned with the digital projection than with the carbon tattoo on the skin surface, according to the surgeon’s postoperative report. This occurred when the tumor was located on the outer side of the breast, and the carbon tattoo marking was performed laterally rather than vertically (Figure 3a,c). For the patient with a large tumor, multiple carbon tattoo markings were performed due to the larger spatial extent of the tumor. (Figure 3b). Table 2. DT alignment results. Deviation between the physical and digital nipple to measure DT alignment, and between the tumor skin projection and the carbon tattoo marking. Observations report accuracy considering intraoperative findings. Patient Nipple deviation Tumor deviation Observations 1 0 mm 10 mm Tumor closer to DT than tattoo 2 1 mm N/A Tumor marked with 4 points; DT aligned with carbon tattoo markings 3 0 mm Large (>30 mm) Tumor closer to DT than tattoo 0 mm N/A 4 5 mm 15 mm N/A Questionnaires After each surgery, the surgeon answered usability questionnaires, including System Usability Score and NASA-TLX [20-23]. The SUS score (1-100 scale) was 95 (range 92.5-10), rated as “Best Imaginable” in the adjective scale [23]. The NASA-TLX scores revealed a very high performance (score=100; range 100-100) with medium mental (score=10; range 10-10), physical (score=10; range 10-10), and temporal (score=10, range 10-20) demands, somewhat high effort (score=20; range 10-90) and medium frustration (score=10; range 10-20). Satisfaction questionnaires reported a preference for the DT-guided XR visualization tool for tumor localization compared to traditional methods. The surgeons rated as excellent the visualization of the tumor, speed of tumor marking, intuitivity of the user interface, gesture recognition sensitivity, and depth perception of the digital contents. Alignment between the DT and the patient was rated fair in the first procedure, good in the second, and excellent in the final two. Latency in the DT display was rated fair in the first procedure and excellent in the subsequent three. Discussion Integrating breast DTs into XR environments marks the next frontier in surgical planning and outcome optimization in BCS. By aligning a patient-specific DT with the patient in the OR and visualizing it through an XR headset, surgeons can see the tumor overlaid on the patient’s body, enabling immediate and precise tumor localization [12,15]. To the best of our knowledge, we present here the first clinically integrated DT-guided XR system for breast tumor localization, achieving fully automated registration without disrupting surgical workflow (Figure 3). This contrasts with conventional tumor localization, such as wire-guided localization, carbon tattooing, magnetics, seeds, or technetium, which are invasive and often associated with high rates of positive margins and re-excision [4,5]. Our results suggest that DTs have the potential to fill the current gap in truly non-invasive methods, which remain absent from clinical practice. Several previous studies have explored XR-based navigation systems for breast tumor localization [17-19, 24]. Recently, breast DTs, created using supine MRI, were aligned to the patient using color and depth sensors, during a breast conservative surgery [18,19]. However, the visualization in these systems was limited to 2D displays, which do not capture the volumetric complexities of breast anatomy. This was attributed to the complexity of integrating 3D model registration into surgical workflows. The system described by Sharifian et al., did not acquire images in real-time; instead, the camera was rotated around the patient and three frames from the sequence were selected to register the AR contents, and then were displayed to the surgeon [18]. The slow nature of this approach limits compatibility with intraoperative workflows and scalability. Ock et al. reported accurate results on tumor localization for BCS performed on a mannequin, using a system designed for a smartphone and performing the DT alignment by detecting the 2D breast phantom’s boundary line [24]. Outside BC, XR systems have been clinically validated in procedures such as open pancreatic surgery and laparoscopic microwave ablation of hepatic hemangiomas [25, 26]. Both studies rely on manual or marker-based registration of static DTs, leading to a decline in accuracy over time due to intraoperative tissue deformation. In BCS, the lack of anatomical reference points in the breast further limits marker-based registration. Although markerless registration methods have been a focus of research, they currently underperform marker-based approaches, limiting their use in real-time clinical settings [27]. In this work, DTs were created from supine MRI, automatically aligned to the patient in real-time using a dual-Azure Kinect 3D sensing system and visualized intraoperatively through an XR headset. Although limited by a small sample size (n=4), tumors varied considerably in volume (0.02–12.74 cm³), and included unifocal and multifocal cases, providing preliminary evidence across heterogeneous presentations. All procedures were performed by an experienced breast cancer surgeon who is also a co-author of this manuscript. Our system accurately aligned the breast DTs with the patients’ breasts with deviations in the nipple position ranging between 0 and 5 mm. In order to achieve this type of alignment, patients were positioned with the arms along the body to more closely resemble the position during the MRI supine acquisition. In one case (patient #4), however, there were visible differences between the breast DT and the patient’s breast shape in the OR, leading to the largest observed nipple deviation (5mm). This can be explained by slight changes between the patient’s position during the MRI acquisition and at the OR. Despite the overall quality of the DT alignment, the tumor projection on the skin did not necessarily correspond to the carbon tattoo marking. This was particularly evident in the most lateral tumor of patient #3, which showed a very large deviation between the vertical projection of the digital tumor and the carbon tattoo marking (>30mm). However, according to the surgeon, the tumor localization intraoperatively matched the one marked by the DT. This highlights the limitations of carbon tattooing, or other conventional methods such as wire-guided, as a reference standard, given their lack of standardization and non-vertical trajectory to the tumor. Accordingly, surgeon confirmation of intraoperative findings remains essential, and margin status will be required for future quantitative validation. The system setup time was 218 seconds (range 72-254) and tumor marking time was 49 seconds (range 32-72) during the procedures, and the surgeon reported these results as compatible with routine surgical workflow. The system was removed within seconds after use, allowing the surgery to proceed with the surgery seamlessly. Usability testing demonstrated high surgeon satisfaction (SUS score=95; range 92.5-100; rated as "Best Imaginable"), and low perceived workload, likely reflecting close co-design with clinicians and workflow prioritization. The variability in effort scores (NASA score=20; range 10-90) may reflect inconsistent cognitive demands across different tumor sizes and anatomy, requiring further investigation in larger cohorts. Progressive improvements in alignment accuracy and display latency suggest a learning curve for both surgeons and the technical team. This collaboration exemplifies a proactive step toward the emerging concept of a medical metaverse in the OR [28-30]. The conceptual distinction between 3D and 4D surgery is critical for fully leveraging DTs. 3D static DT are reconstructed from different imaging modalities to enhance personalized surgical planning, tumor segmentation, and preoperative simulations. 4D models can extend this framework by incorporating temporal dynamics, such as respiratory motion, tissue deformation, and real-time navigation from intraoperative sensors, which are critical for XR-based navigation systems in the OR [30]. In contrast, conventional surgery planning workflows rely mainly on reports and 2D imaging analysis, imposing a high-cognitive burden on surgeons. They must keep a memory of preoperative medical images, infer tumor location on the patient's breast despite differences between imaging and surgical positioning, while cognitively performing breast and tumor volumetric assessment. In this work, shadow twins are overlaid on the patient in continuous 15 fps surface scanning to track respiratory motion, supporting more accurate intraoperative decision-making. This system represents a concrete implementation of 4D surgery principles rather than just a preliminary step. Besides providing high value to intraoperative planning, incorporating DTs into the preoperative planning workflow also enables patients to receive further validation of their concerns through treatment explanations and postoperative outcomes [31,32]. Thus, patients can be better educated and participate more proactively in the surgery decision-making process. Simultaneously, DTs can further serve as a ground-truth to validate other emerging imaging techniques, such as AI-driven 3D ultrasound [33]. Future work should extend this proof-of-concept to a larger patient cohort, assess accuracy using margin status as the primary outcome measure, and include a larger group of surgeons including some with limited XR experience. Additionally, advances in biomechanical breast modeling and non-rigid registration are needed to accurately simulate tissue deformation and tumor displacement across different patient positions and throughout breathing motion [30,34-36]. Another current limitation is the reliance on manual segmentation by a physician with radiologist validation, which constrains scalability. Incorporating automatic segmentation of MRI images could streamline model generation, reduce operator dependency, and facilitate broader clinical adoption [37-39]. Beyond technical improvements, ethical considerations, social impact, and legal guidance remain critical considerations for safe and responsible clinical implementation [12]. Such improvements could further enhance the predictive power and clinical value of DTs in breast cancer surgery. For translation into clinical practice, compliance with the European Union Medical Device Regulation (MDR, Regulation (EU) 2017/745) will be required, including conformity assessment, clinical evaluation, and Conformité Européenne (CE) marking to ensure safety and performance. By integrating DTs in XR environments, new paradigms of interaction with patient-specific data are available to surgeons and healthcare professionals overall, contributing to non-invasive and intraoperative tumor localization, with digital precision. This proof-of-concept demonstrates the feasibility of automatically aligning DTs with patients in the OR while meeting clinical workflow requirements, enabling accurate tumor localization, and enhanced surgical performance. Methods Ethical approval Patients were enrolled in the study following study protocol presentation and signed informed consent. The study was performed in compliance with relevant laws and institutional guidelines and was approved by the ethics committee of Champalimaud Foundation on March 2022 with the identification Breast 4.0 v 3.0 (no applicable approval number). Patient-specific Digital Twins for Breast Cancer Surgery Four patients were selected and consented for acquisition of contrast-enhanced MRI images. Contrast-enhanced MRI images were taken after the administration of a contrast agent to improve lesion visualization. An additional scan in the supine position was acquired after the standard prone sequence to build the DTs with the patient in a position similar to the one in the OR (with the arms along the body). All scans were performed on a 3T MR system, using a dedicated 16-channel breast coil (for the prone acquisition) and for the supine acquisition a 32-channel torso coil was used together with a position device that allows the usage of the coil without distorting the anatomy of the patient. The supine acquisition was a 3D T1-weighted Dixon scan of the thorax with breathing compensation and the following scan parameters: TE=1.57 ms; TR=5.3 ms; FOV=450x450 mm^2, slices=250; in-plane resolution=1.2x1.2 mm^2; slice thickness=1.5 mm; and a total scan time=05:52 s. Skin, fibroglandular tissue, and tumor segmentations were performed on the supine MRI images using 3D Slicer by a physician and validated by an experienced radiologist (Figure 2). The patient’s image data was anonymized, stored in a research database, and used to create a static DT of the patient. An additional scan of the patient’s skin surface was acquired using the Go!Scan 50 handheld scanning device, in supine position, for validation and prior testing of the registration algorithm before surgery. Post segmentation, the 3D model was exported and modelled using the Open3D library in Python. Since the segmentation output contained the whole volume of the breast (Figure 4a), the model’s surface was first extracted. By default, 10,000 rays were cast from the outside to the center of the mesh, with each collision resulting in a vertice on the final mesh (Figure 4b). Rays near the poles were removed to avoid intersections with points inside the breast volume. Statistical outliers were removed by calculating the mean distance between each point and its neighboring points. The global mean and standard deviation of these average distances were computed with points with an average distance greater than mean + (std_ratio × standard deviation) being classified as outliers and excluded from further analysis (Figure 4c). The surface was clipped along the sagittal axis to remove areas on the side and back. Optimized XR interface The DT was uploaded to the XR application and visualized through the ML. Co-design sessions were conducted with surgeons in the OR to inform the XR interface design. Early concepts were implemented in Unity 3D with MRTK3 and iteratively based on surgeons’ feedback (Figure 5a-b). This process focused on optimizing tumor visualization for accurate marking, leading to the adoption of visual cues, such as a skin grid and tumor projection onto the surface (Figure 6 and Supplementary video). The interface menu was spatially anchored to the DT, constrained to move within a viewing frustum defined relative to the DT’s coordinate system, oriented toward the surgeon (Figure 5c and Supplementary video). Positioned at an approximate viewing distance of 60 cm from the surgeon, the interface supported head-gaze targeting and pinch-based selection (Figure 5c-d). Throughout iterative refinement, surgeon feedback and prototype development converged towards a high-fidelity interface suitable for use in the OR. Surgeons emphasized the importance of visualizing the tumor projection on the skin, given the lack of standardization in conventional carbon tattooing. Tumor marking was defined as a vertical vector from the tumor center to the skin surface, parallel to gravity’s vector (Figure 6a-b). The projection was considered helpful during tumor marking. Additionally, the ML’s near clipping boundary (~37 cm) was intentionally used as a visual aid to better understand tumor depth and position (Figure 6c-d and supplementary video). This provided an easy way for the surgeon to switch the visualization of the digital content on or off by moving further or closer to the patient. 3D sensing system for DT to patient-alignment To align the patient’s breast DT to the patient in real-time in the OR, a 3D surface scan system composed of two Azure Kinect cameras (Figure 7a) was developed and validated with patient data [40]. The system was designed to be attached to the operating table and incorporated placeholders for AprilTag markers. These markers establish a common coordinate system between the Azure Kinects and the ML headset. Upon patient detection by the 3D surface scan system, the DT is automatically registered to the patient using an interactive closest point (ICP) algorithm and displayed in the XR interface (Figure 7b). The surface scanning system continually updates the DT position (15 fps) displayed to the surgeon through the ML, to keep up with the patient’s breathing motion. In this configuration, the DT operates not as a static model but as a dynamic shadow twin, integrating real-time data to maintain accurate alignment with the patient [11]. Then, the surgeon selects the option to freeze the DT position at the instant of the marking task, to ensure its stability during those seconds. All procedures were performed by a breast surgeon with over 15 years of experience in breast-conserving surgery, who is a co-author of this study. The surgeon had prior experience with the XR system before clinical implementation. Declarations Data availability The breast supine MRIs, tissue segmentations, and DTs for the four patients are available in this study repository, LINK_TO_REPOSITORY. Code availability The underlying code for the XR system and DTs alignment in the OR is not publicly available for proprietary reasons. Acknowledgements This research is part of the Health for Portugal (HfP), funded by Agendas Mobilizadoras para a Inovação Empresarial - Plano de Recuperação e Resiliência (PRR) Português (no applicable grant number). The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. The authors would like to thank the Breast Unit at Champalimaud Foundation for their cooperation in this study. Author contributions RT, JS, TM, and PG conceptualized and designed the study. RT acquired the patients’ surface scan prior to surgery. AC and MC performed the breast segmentations. NL, YF, and JS managed the medical imaging processing pipeline. BV, AL, and TM designed and implemented the DTs production pipeline and the XR system for the OR. RT, DSL, and TM led the user interface design. RT and TM analyzed and interpreted the data. RT, TM, and PG prepared the original draft. All authors reviewed and edited the manuscript and read and approved the final version. Competing Interests Authors JS, TM, and PG have a provisional patent application - METHODS AND SYSTEMS FOR PRECISION-GUIDED SURGERY - pending to Champalimaud Foundation but declare no other competing interests. All other authors declare no financial or non-financial competing interests. References The International Agency for Research on Cancer (IARC). (n.d.). Global Cancer Observatory . https://gco.iarc.who.int/en. Last accessed 2024/02/08. Cardoso, F., Kyriakides, S., Ohno, S., Penault-Llorca, F., Poortmans, P., Rubio, I. T., ... & Senkus, E. (2019). Early breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Annals of oncology , 30 (8), 1194-1220. Yu, C. X. (2023). Radiotherapy of early‐stage breast cancer. Precision Radiation Oncology , 7 (1), 67-79. Corsi F, Sorrentino L, Bossi D, Sartani A, Foschi D. Preoperative localization and surgical margins in conservative breast surgery. Int J Surg Oncol 2013;2013: 793819. Ryan, J. F., Ghosh, S., Rajaee, A. N., Olson, D., Lesniak, D. M., Peiris, L. J., & Alberta Breast Surgical Oncology Research Group (2024). Re-excision rates after breast-conserving surgery for invasive breast cancer: an Albertan perspective. Canadian journal of surgery. Journal canadien de chirurgie, 67(6), E363–E369. https://doi.org/10.1503/cjs.004723 W. L. Richey, J. S. Heiselman, M. J. Ringel, I. M. Meszoely and M. I. Miga, "Computational Imaging to Compensate for Soft-Tissue Deformations in Image-Guided Breast Conserving Surgery," in IEEE Transactions on Biomedical Engineering , vol. 69, no. 12, pp. 3760-3771, Dec. 2022, doi: 10.1109/TBME.2022.3177044. Bessa, S., Gouveia, P. F., Carvalho, P. H., Rodrigues, C., Silva, N. L., Cardoso, F., ... & Cardoso, M. J. (2020). 3D digital breast cancer models with multimodal fusion algorithms. The Breast , 49 , 281-290. Garzotto, F., Comoretto, R. I., Michieletto, S., Franzoso, G., Mele, M. L., Gregori, D., ... & Saibene, T. (2021). Preoperative non-palpable breast lesion localization, innovative techniques and clinical outcomes in surgical practice: A systematic review and meta-analysis. The Breast , 58 , 93-105. Barth, R. J., Jr, Krishnaswamy, V., Paulsen, K. D., Rooney, T. B., Wells, W. A., Angeles, C. V., Zuurbier, R. A., Rosenkranz, K., Poplack, S., & Tosteson, T. D. (2019). A Randomized Prospective Trial of Supine MRI-Guided Versus Wire-Localized Lumpectomy for Breast Cancer. Annals of surgical oncology , 26 (10),3099–3108. https://doi.org/10.1245/s10434-019-07531-4 Fausto, A., Fanizzi, A., Volterrani, L., Mazzei, F. G., Calabrese, C., Casella, D., Marcasciano, M., Massafra, R., La Forgia, D., & Mazzei, M. A. (2020). Feasibility, Image Quality and Clinical Evaluation of Contrast-Enhanced Breast MRI Performed in a Supine Position Compared to the Standard Prone Position. Cancers , 12 (9), 2364. https://doi.org/10.3390/cancers12092364 Zhang, K., Zhou, H. Y., Baptista-Hon, D. T., Gao, Y., Liu, X., Oermann, E., ... & Wu, J. (2024). Concepts and applications of digital twins in healthcare and medicine. Patterns , 5 (8). Mekki, Y.M., Luijten, G., Hagert, E. et al. Digital twins for the era of personalized surgery. npj Digit. Med. 8, 283 (2025). https://doi.org/10.1038/s41746-025-01575-5. Shi, Y., Deng, X., Tong, Y., Li, R., Zhang, Y., Ren, L., & Si, W. (2022). Synergistic digital twin and holographic augmented-reality-guided percutaneous puncture of respiratory liver tumor. IEEE Transactions on Human-Machine Systems , 52 (6), 1364-1374. Kolekar, A., Shalgar, S., & Malawade, I. (2023, September). Beyond reality: A study of integrating digital twins. In Journal of Physics: Conference Series (Vol. 2601, No. 1, p. 012030). IOP Publishing. Katsoulakis, E., Wang, Q., Wu, H. et al. Digital twins for health: a scoping review. npj Digit. Med. 7, 77 (2024). https://doi.org/10.1038/s41746-024-01073-0. Asciak, L., Kyeremeh, J., Luo, X. et al. Digital twin assisted surgery, concept, opportunities, and challenges. npj Digit. Med. 8, 32 (2025). https://doi.org/10.1038/s41746-024-01413-0. Gouveia, P. F., Costa, J., Morgado, P., Kates, R., Pinto, D., Mavioso, C., ... & Cardoso, M. J. (2021). Breast cancer surgery with augmented reality. The Breast , 56 , 14-17. Sharifian, R., Madad-Zadeh, S., Bourdel, N., Giro, A., Marraoui, W., Pomel, C., & Bartoli, A. (2024, October). Guiding Breast Conservative Surgery by Augmented Reality from Preoperative MRI: Initial System Design and Retrospective Trials. In Deep Breast Workshop on AI and Imaging for Diagnostic and Treatment Challenges in Breast Care (pp. 212-220). Cham: Springer Nature Switzerland. Lee, M., Woo, J., Peak, S. H., Kim, H. G., Lim, W. S., Chung, J., ... & Lee, J. W. (2024). An exploratory clinical trial of preoperative non-invasive localization before breast-conserving surgery using augmented reality technology. Breast Cancer Research and Treatment , 206 (1), 31-44. Hart, S. G., & Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In Advances in psychology (Vol. 52, pp. 139-183). North-Holland. Byers, J. C., Bittner, A. C., & Hill, S. G. (1989). Traditional and raw task load index (TLX) correlations: Are paired comparisons necessary. Advances in industrial ergonomics and safety, 1, 481-485. Brooke, J. (1996). SUS-A quick and dirty usability scale. Usability evaluation in industry, 189(194), 4-7. Bangor, A., Kortum, P., & Miller, J. (2009). Determining what individual SUS scores mean: Adding an adjective rating scale. Journal of usability studies, 4(3), 114-123. Ock, J., Moon, S., Kim, M., Ko, B. S., & Kim, N. (2024). Evaluation of the accuracy of an augmented reality-based tumor-targeting guide for breast-conserving surgery. Computer Methods and Programs in Biomedicine , 245 , 108002. Lan, T., Liu, S., Dai, Y., Luo, J., Han, J., & Jin, Y. (2025). Real-time image fusion and Apple Vision Pro in laparoscopic microwave ablation of hepatic hemangiomas. NPJ Precision Oncology , 9 (1), 79. Javaheri, H., Ghamarnejad, O., Lukowicz, P., Stavrou, G. A., & Karolus, J. (2025, April). From Concept to Clinic: Multidisciplinary Design, Development, and Clinical Validation of Augmented Reality-Assisted Open Pancreatic Surgery. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (pp. 1-24). M. Tu, H. Jung, A. Moghadam, J. Raythatha, J. Hsu and J. Kim, "Exploring the Performance of Geometry-Based Markerless Registration in a Simulated Surgical Environment: A Comparative Study of Registration Algorithms in Medical Augmented Reality," 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) , Sydney, Australia, 2023, pp. 1-4, doi: 10.1109/EMBC40787.2023.10341197. Navab, N., Martin-Gomez, A., Seibold, M., Sommersperger, M., Song, T., Winkler, A., ... & Eck, U. (2022). Medical augmented reality: definition, principle components, domain modeling, and design-development-validation process. Journal of Imaging , 9 (1), 4. Gouveia, P. F., Luna, R., Fontes, F., Pinto, D., Mavioso, C., Anacleto, J., ... & Cardoso, M. J. (2023). Augmented reality in breast surgery education. Breast Care , 18 (3), 182-186. Riedel, P., Riesner, M., Wendt, K., & Aßmann, U. (2022, May). Data-driven digital twins in surgery utilizing augmented reality and machine learning. In 2022 IEEE International Conference on Communications Workshops (ICC Workshops) (pp. 580-585). IEEE. Kaidar-Person, O., Antunes, M., Cardoso, J. S., Ciani, O., Cruz, H., Di Micco, R., ... & CINDERELLA Consortium. (2023). Evaluating the ability of an artificial-intelligence cloud-based platform designed to provide information prior to locoregional therapy for breast cancer in improving patient’s satisfaction with therapy: The CINDERELLA trial. PLoS One , 18 (8), e0289365. Bonci, E. A., Kaidar-Person, O., Antunes, M., Ciani, O., Cruz, H., Di Micco, R., ... & Cardoso, M. J. (2024). CINDERELLA Trial: validation of an artificial-intelligence cloud-based platform to improve the shared decision-making process and outcomes in breast cancer patients proposed for locoregional treatment. European Journal of Surgical Oncology , 50 (2). Kwon, H., Oh, S. H., Kim, M. G., Kim, Y., Jung, G., Lee, H. J., ... & Bae, H. M. (2024). Enhancing Breast Cancer Detection through Advanced AI-Driven Ultrasound Technology: A Comprehensive Evaluation of Vis-BUS. Diagnostics , 14 (17), 1867. Carolina, L., Timoteo, R., Laborde, A., Vaz, B., Gouveia, P., Marques, T. (2025). Volumetric Correction for Aligning Breast Digital Twins with Surface Scan. Proceedings of the 14th Conference on New Technologies for Computer/Robot Assisted Surgery. Teixeira, A. M., & Martins, P. (2023). A review of bioengineering techniques applied to breast tissue: Mechanical properties, tissue engineering and finite element analysis. Frontiers in bioengineering and biotechnology , 11 , 1161815. https://doi.org/10.3389/fbioe.2023.1161815 Eiben, B., Vavourakis, V., Hipwell, J. H., Kabus, S., Buelow, T., Lorenz, C., Mertzanidou, T., Reis, S., Williams, N. R., Keshtgar, M., & Hawkes, D. J. (2016). Symmetric Biomechanically Guided Prone-to-Supine Breast Image Registration. Annals of biomedical engineering, 44(1), 154–173. https://doi.org/10.1007/s10439-015-1496-z Forghani, Y., Timóteo, R., Marques, T., Loução, N., Cardoso, M. J., Cardoso, F., ... & Santinha, J. (2025). Comparative analysis of nnU-Net and Auto3Dseg for fat and fibroglandular tissue segmentation in MRI. Journal of Medical Imaging , 12 (2), 024005-024005. Lew, C.O., Harouni, M., Kirksey, E.R. et al. A publicly available deep learning model and dataset for segmentation of breast, fibroglandular tissue, and vessels in breast MRI. Sci Rep 14, 5383 (2024). https://doi.org/10.1038/s41598-024-54048-2 Narimani, S., Roth Hoff, S., Dæhli Kurz, K. et al. Comparative analysis of deep learning architectures for breast region segmentation with a novel breast boundary proposal. Sci Rep 15, 8806 (2025). https://doi.org/10.1038/s41598-025-92863-3 Timoteo, R., Laborde, A., Forghani, Y., Lopes, D.S., Gouveia, P., Marques, T. (2025). Automatic Surface Scan System for Breast Cancer Surgery. Proceedings of the 14th Conference on New Technologies for Computer/Robot Assisted Surgery . Additional Declarations Competing interest reported. Authors JS, TM, and PG have a provisional patent application - METHODS AND SYSTEMS FOR PRECISION-GUIDED SURGERY - pending to Champalimaud Foundation but declare no other competing interests. All other authors declare no financial or non-financial competing interests. Supplementary Files Streamingpatient4.mp4 Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8907845","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":595081594,"identity":"c254aa41-1da0-4a79-a13c-de5539df6b46","order_by":0,"name":"Rafaela Timóteo","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Rafaela","middleName":"","lastName":"Timóteo","suffix":""},{"id":595081595,"identity":"4813560c-4073-4232-9092-517309fd7c38","order_by":1,"name":"Alexandre Laborde","email":"","orcid":"","institution":"IT People","correspondingAuthor":false,"prefix":"","firstName":"Alexandre","middleName":"","lastName":"Laborde","suffix":""},{"id":595081596,"identity":"30e86564-bab6-4e07-bfbd-80e150025f8f","order_by":2,"name":"Bruno Vaz","email":"","orcid":"","institution":"IT People","correspondingAuthor":false,"prefix":"","firstName":"Bruno","middleName":"","lastName":"Vaz","suffix":""},{"id":595081597,"identity":"5dfd0534-b9ef-4c42-8bc8-c560cf4357bf","order_by":3,"name":"Nuno Loução","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Nuno","middleName":"","lastName":"Loução","suffix":""},{"id":595081598,"identity":"5a81938c-e4e5-4b9b-bac5-6b8f4d372c8e","order_by":4,"name":"Yasna Forghani","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Yasna","middleName":"","lastName":"Forghani","suffix":""},{"id":595081599,"identity":"061ad929-9c9e-41e1-897a-4da724d34a45","order_by":5,"name":"André Cardoso","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"André","middleName":"","lastName":"Cardoso","suffix":""},{"id":595081600,"identity":"f2ab9f8d-e609-48ed-b6e5-c423384d2c01","order_by":6,"name":"Mariana Correia","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Mariana","middleName":"","lastName":"Correia","suffix":""},{"id":595081601,"identity":"5db84611-8156-41ab-8e78-8bba4731b145","order_by":7,"name":"João Santinha","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"João","middleName":"","lastName":"Santinha","suffix":""},{"id":595081602,"identity":"92d32626-8bd3-47f1-91ff-9418d21812a3","order_by":8,"name":"Daniel Simões Lopes","email":"","orcid":"","institution":"University of Lisbon","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"Simões","lastName":"Lopes","suffix":""},{"id":595081603,"identity":"53cbd9a5-1f09-4e45-aab7-8e4d9ecac020","order_by":9,"name":"Tiago Marques","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABgElEQVRIie2STUsCQRiA30XQy4jXN4z8BcHKwGRg7l8Z2UOXPQgL4kFqIVhP1nUlyb+gCNIxCPQy0HVhQgjJOnjYCKS9RLt+sukPCNrnMJ/vM1/vAMTE/EUSkV5tVSsWZAAXzQIJS9zGkKgitsqBtQzDHSViKPYeBX4pWioxefdgkDtpGPkX/258lGtcTT3/HjBze/Xar9Tx8Lgh8i7UxiVrdbAkO3VA5ttiRml6YFJVDGmrKQDxecikM0TChEELIEx9rQCjBKTioMGyyoCXu8hpIjjhhYqcSZIMlAeDoWJzfX391DxUtEA58f02v+x03j5DBVU8n0vyHShPs6hC6CRQyuEukLY4B5fQlWIwmbYDxV3uUlopj6SqOKrUHTI1s2TI811hmErTRkTXqMrWdahMzQIXwWoLUjejvufV5JmT0nsfX3WeyzVGffDtImac84GszIsae9J7rlfj2uadkwhqWPNIipaJ2CY6mC1b607C2wzuEPkb2p6AmJiYmP/AD2K/hjnNi6AwAAAAAElFTkSuQmCC","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":true,"prefix":"","firstName":"Tiago","middleName":"","lastName":"Marques","suffix":""},{"id":595081604,"identity":"09c6954d-7e0f-4777-88d1-faaf13d45fe4","order_by":10,"name":"Pedro Gouveia","email":"","orcid":"","institution":"Champalimaud Foundation","correspondingAuthor":false,"prefix":"","firstName":"Pedro","middleName":"","lastName":"Gouveia","suffix":""}],"badges":[],"createdAt":"2026-02-18 09:38:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8907845/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8907845/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103567705,"identity":"3b0a5dae-d79d-4a42-b195-2b1d0256820f","added_by":"auto","created_at":"2026-02-27 07:35:24","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":160316,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOverview of the system pipeline: from Digital Twin creation to XR display in the OR. \u003c/strong\u003eBreast tissues were segmented from supine MRI scans using 3D Slicer, to generate breast DTs. Afterwards, the DT was modelled using the Open3D library in Python and custom-made software. In the OR, the DTs were registered to real-time surface scans using two Microsoft Azure Kinect sensors and displayed to the surgeon using an XR HMD (Magic Leap 2). The interface visualized through the XR headset was implemented using MRTK3 and Unity 3D.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/0738213d1f30d51a998ce8d5.png"},{"id":103567712,"identity":"bfb04a30-38d8-46da-b0cf-e81942f48ff9","added_by":"auto","created_at":"2026-02-27 07:35:25","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2259046,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDigital Twins (DTs) were created from supine breast MRI.\u003c/strong\u003e a) Raw supine MRI images for the four patients in the study (one in each row). b) Segmented supine MRI images: fat (blue), fibroglandular tissue (white), and lesions (pink). c) 3D visualization of the segmentation of the MRI images, corresponding to the static DT: fat (white/grey), fibroglandular tissue (blue), and lesions (red).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/c3526d7de108e2c8160e75fc.png"},{"id":103567706,"identity":"51b8c63d-0569-4892-95be-5f23adf25a8d","added_by":"auto","created_at":"2026-02-27 07:35:25","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1064908,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDT-guided tumor localization in the operating room (OR). \u003c/strong\u003eImages taken from live captures performed through the XR headset (ML) of the patients’ DTs aligned with the torso in the OR for the four surgeries (a-d). Fibroglandular tissue is shown in blue, the tumor in red, the nipple in yellow, the tumor’s vertical projection in pink (a line and circle on the skin’s surface), and a grid to help perceive distances in white. In d) the image was taken in a moment when the grid was not being displayed.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/b0aceb6b52709888465f731f.png"},{"id":104398313,"identity":"0f7f1243-60e9-419b-bb31-a306d4cdcecd","added_by":"auto","created_at":"2026-03-11 12:01:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":392596,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDT modelling for surface extraction.\u003c/strong\u003e a) MRI segmentation output. b) DT surface extraction technique - points surrounding the DT from which rays were cast from the outside to the center of the mesh, resulting in collisions that determined the points to constitute the final mesh. The color of the points represents the polar angle with extremes being removed. c) Removal of statistical outliers (points in red) after surface extraction.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/03281593ee3bf6d4fc74d33e.png"},{"id":104398170,"identity":"c79cf102-be64-4099-8f79-6f01c0a945f9","added_by":"auto","created_at":"2026-03-11 12:00:10","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1481700,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eXR interface development with co-design sessions. \u003c/strong\u003ea) Patient DT, including fibroglandular tissue (in blue), tumor (in red), tumor projection (in pink), and a 1 cm spaced grid (in white). b) User interface menu first prototypes design after the requirements meeting. c) Final user interface with head-gaze and pinch gesture to select buttons. d) Surgeon interacting with the user interface using the head-gaze and pinch gesture.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/4d37ca856731dec5dbae82bc.png"},{"id":103567708,"identity":"802e56cd-7896-47b4-985a-05e10d4cdb9a","added_by":"auto","created_at":"2026-02-27 07:35:25","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1761319,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDT visualization and visual guides. \u003c/strong\u003ea) Visualization of the DT in the final chosen color scheme and including visual guides. b) Side view of the tumor vertical projection on the breast - a pink line going from the center of the lesion to the skin surface, with the surface collision displayed as a pink circle. c) DT visualized through the ML aligned with the patient. d) DT visualization through the ML when closer than the minimum near boundary, resulting in a clipping of the contents.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/24d94b23d8aadfda5d1c859a.png"},{"id":103567710,"identity":"aee8b9c8-554e-490b-889d-4020cc1b871c","added_by":"auto","created_at":"2026-02-27 07:35:25","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1583889,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDT-based XR navigation system for breast cancer surgery. \u003c/strong\u003ea) System setup in the OR. b) Workflow pipeline of the alignment of a patient’s DT to the surface scan acquired in real-time in the OR, and its display through the XR headset. c) Alignment of the DT (in green) to the patient’s body surface (pink). d) DT visualized through the ML in the OR.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/21228d776309467bc777dfad.png"},{"id":104407404,"identity":"c1e61a4b-245d-452d-968d-4237bb316e82","added_by":"auto","created_at":"2026-03-11 12:37:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":9115043,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/ca92f50f-63a2-49f9-9607-cf245f7b7b19.pdf"},{"id":103567713,"identity":"7c88403f-1c1e-476d-8af1-30798863f307","added_by":"auto","created_at":"2026-02-27 07:35:26","extension":"mp4","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":63234753,"visible":true,"origin":"","legend":"","description":"","filename":"Streamingpatient4.mp4","url":"https://assets-eu.researchsquare.com/files/rs-8907845/v1/5b4bebdc4be7da23b3e066de.mp4"}],"financialInterests":"Competing interest reported. Authors JS, TM, and PG have a provisional patent application - METHODS AND SYSTEMS FOR PRECISION-GUIDED SURGERY - pending to Champalimaud Foundation but declare no other competing interests. All other authors declare no financial or non-financial competing interests.","formattedTitle":"Digital twins for breast cancer surgery","fulltext":[{"header":"Introduction","content":"\u003cp\u003eBreast cancer (BC) is the second most common cancer worldwide, with approximately 2.3 million new cases reported annually [1,2]. In Western countries, the widespread implementation of screening programs has increased the rate of early diagnosis. As a result, breast-conserving surgery (BCS) followed by adjuvant radiotherapy has become the gold standard for local treatment, accounting for 60\u0026ndash;80% of newly diagnosed cases in 2022 [3]. Early detection often involves small, non-palpable lesions, which complicates surgical management. Successful BCS relies on precise tumor localization and complete excision with clear margins to minimize the risk of re-excision [4,5]. Surgical planning, decision-making, and outcomes rely on imaging modalities, with the most common being breast magnetic resonance imaging (MRI), acquired in the prone position. Due to the deformable nature of the breast tissue, and position change between imaging and surgery, planning is particularly challenging [6,7]. Localization of non-palpable lesions requires invasive techniques, including wire-guided, carbon tattooing, biopsy clips,\u0026nbsp;radioactive seed localization,\u0026nbsp;radio-occult lesion localization, and magnetic seeds [4]. In recent years, non-wired, non-ionizing devices have emerged [8], and supine MRI has been proposed as a solution to better approximate the patient\u0026rsquo;s surgical position [9,10]. Despite these advances, conventional tumor localization approaches remain percutaneous, often preoperative, uncomfortable for patients, and lacking optimal precision.\u003c/p\u003e\n\u003cp\u003eDigital twins (DTs) are emerging in healthcare as patient-specific dynamic digital replicas, integrating both anatomical and physiological information, as a powerful tool for screening, diagnosis, treatment, decision-making, simulations, and outcome predictions [11, 12]. Generated from medical imaging, such as MRI, DTs can be aligned with the anatomy in the operating room (OR) and visualized through extended reality (XR) head-mounted displays (HMDs) to guide interventions [13,14]. Integrating DTs and XR in surgery offers digital, non-invasive guidance with benefits for both patient satisfaction and surgical performance [15].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDTs can be classified by the data they contain and how frequently they are updated. Most commercial solutions are centered on static twins, which do not change over time [12]. By incorporating finite element algorithms, functional twins can simulate the tissue behavior under specific conditions [12]. Shadow (self-adaptive) twins integrate real-time sensors or imaging data to account for tissue deformation and respiratory motion, optimizing surgical workflows [2,12]. Intelligent twins extend this further by providing continuous feedback and real-time risk prediction, simulations, and surgical planning, often leveraging biometric devices, implantable sensors, and machine learning [11,12]. Despite growing research, intraoperative use of DTs in XR remains largely experimental, with most work to date focused on orthopedics or neurosurgery [12-16]. In breast cancer surgical treatment, studies of such character remain scarce [17-19].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn this study, we present a DT-based XR navigation system for BCS tumor localization in the OR and perform a proof-of-concept pilot study to validate its performance. Breast DTs were created from segmented supine breast MRI images, aligned to the patient\u0026rsquo;s torso using color and depth sensors, and displayed to the surgeon through an XR headset during BCS (Figure 1). This proof-of-concept study provides initial evidence of the feasibility of an XR-based, non-invasive tumor localization in BCS.\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eCohort\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFour (4) breast cancer patients scheduled for breast surgery were recruited via a signed informed consent form at the Breast Unit at Champalimaud Foundation, between July and November 2025. The median patient age was 48 years (range 44-51). In addition to the standard-of-care breast MRI acquisition in prone, patients underwent breast MRI in supine (Figure 2a). Breast supine MRIs were segmented into three tissues: fat, fibroglandular tissue (FGT), and tumor (Figure 2b). One patient had the tumor in the left breast, and the other three in the right breast. Tumor volume and oriented bounding box (OBB) diameters were calculated from the segmentations (Table 1). One patient had a large tumor (maximum diameter of 74.2 mm) and underwent mastectomy followed by breast reconstructive surgery. Another patient had a very small satellite lesion near the primary tumor (Table 1). Breast static DTs were created from the segmented MRI images after post-processing (Figure 2c).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1. Patients\u0026rsquo; lesion description after segmentation.\u0026nbsp;\u003c/strong\u003ePatient #2 had a large lesion and underwent a mastectomy. Patient #3 had a small satellite lesion near the primary tumor. Patient #3 had the tumor on the left breast, while the others had the tumor on the right breast.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"527\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePatient #\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 123px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBreast with tumor\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 137px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor volume (cm\u0026sup3;)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 163px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOBB diameter -\u0026nbsp;\u003cbr\u003e\u0026nbsp;WxHxL (mm)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 123px;\"\u003e\n \u003cp\u003eRight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 137px;\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 163px;\"\u003e\n \u003cp\u003e5.5 x 8.1 x 11.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 123px;\"\u003e\n \u003cp\u003eRight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 137px;\"\u003e\n \u003cp\u003e12.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 163px;\"\u003e\n \u003cp\u003e35.9 x 50.8 x 74.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 104px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 123px;\"\u003e\n \u003cp\u003eLeft\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 137px;\"\u003e\n \u003cp\u003e0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 163px;\"\u003e\n \u003cp\u003e6.5 x 10.9 x 13.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 137px;\"\u003e\n \u003cp\u003e0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 163px;\"\u003e\n \u003cp\u003e4.0 x 4.5 x 5.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 123px;\"\u003e\n \u003cp\u003eRight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 137px;\"\u003e\n \u003cp\u003e1.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 163px;\"\u003e\n \u003cp\u003e14.4 x 15.2 x 21.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTumor marking in the operating room\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe system was set up after patient anesthesia, with the setup time taking 218 seconds (range 72-190). During setup, the surgeon was asked to put on the Magic Leap 2 (ML) headset and complete a standard eye-calibration process. Then, the surgeon entered the XR application and the patient\u0026rsquo;s DT was displayed and aligned within seconds (Figure 3). The visualization mode in XR contained the tumor in red, FGT in blue, nipples in yellow, a 1cm grid along the patient\u0026rsquo;s skin in white, and the tumor projection on the skin in pink. Tumor marking took on average 49 seconds (range 32-72).\u003c/p\u003e\n\u003cp\u003eAcross patients, the digital nipple deviation ranged from 0 mm to 5 mm, indicating a very good alignment between the breast DT and the patient\u0026rsquo;s anatomy in the OR (Figure 3). Tumor skin projection deviations relative to carbon tattoo markings ranged from 0 mm to large deviations (\u0026gt; 30 mm) (Table 2). In two cases where large deviations were detected, the intraoperative tumor position was more vertically aligned with the digital projection than with the carbon tattoo on the skin surface, according to the surgeon\u0026rsquo;s postoperative report. This occurred when the tumor was located on the outer side of the breast, and the carbon tattoo marking was performed laterally rather than vertically (Figure 3a,c). For the patient with a large tumor, multiple carbon tattoo markings were performed due to the larger spatial extent of the tumor. (Figure 3b).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2. DT alignment results.\u0026nbsp;\u003c/strong\u003eDeviation between the physical and digital nipple to measure DT alignment, and between the tumor skin projection and the carbon tattoo marking. Observations report accuracy considering intraoperative findings.\u003c/p\u003e\n\u003cdiv\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"516\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 72px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePatient\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNipple deviation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 122px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor deviation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 211px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eObservations\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 72px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e0 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 122px;\"\u003e\n \u003cp\u003e10 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 211px;\"\u003e\n \u003cp\u003eTumor closer to DT than tattoo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 72px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e1 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 122px;\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 211px;\"\u003e\n \u003cp\u003eTumor marked with 4 points; DT aligned with carbon tattoo markings\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 72px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 111px;\"\u003e\n \u003cp\u003e0 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 122px;\"\u003e\n \u003cp\u003eLarge (\u0026gt;30 mm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 211px;\"\u003e\n \u003cp\u003eTumor closer to DT than tattoo\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 122px;\"\u003e\n \u003cp\u003e0 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 211px;\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 72px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e5 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 122px;\"\u003e\n \u003cp\u003e15 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 211px;\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuestionnaires\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter each surgery, the surgeon answered usability questionnaires, including System Usability Score and NASA-TLX [20-23]. The SUS score (1-100 scale) was 95 (range 92.5-10), rated as \u0026ldquo;Best Imaginable\u0026rdquo; in the adjective scale [23]. The NASA-TLX scores revealed a very high performance (score=100; range 100-100) with medium mental (score=10; range 10-10), physical (score=10; range 10-10), and temporal (score=10, range 10-20) demands, somewhat high effort (score=20; range 10-90) and medium frustration (score=10; range 10-20).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSatisfaction questionnaires reported a preference for the DT-guided XR visualization tool for tumor localization compared to traditional methods. The surgeons rated as excellent the visualization of the tumor, speed of tumor marking, intuitivity of the user interface, gesture recognition sensitivity, and depth perception of the digital contents. Alignment between the DT and the patient was rated fair in the first procedure, good in the second, and excellent in the final two. Latency in the DT display was rated fair in the first procedure and excellent in the subsequent three.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIntegrating breast DTs into XR environments marks the next frontier in surgical planning and outcome optimization in BCS. By aligning a patient-specific DT with the patient in the OR and visualizing it through an XR headset, surgeons can see the tumor overlaid on the patient\u0026rsquo;s body, enabling immediate and precise tumor localization [12,15]. To the best of our knowledge, we present here the first clinically integrated DT-guided XR system for breast tumor localization, achieving fully automated registration without disrupting surgical workflow (Figure 3). This contrasts with conventional tumor localization, such as wire-guided localization, carbon tattooing, magnetics, seeds, or technetium, which are invasive and often associated with high rates of positive margins and re-excision [4,5]. Our results suggest that DTs have the potential to fill the current gap in truly non-invasive methods, which remain absent from clinical practice.\u003c/p\u003e\n\u003cp\u003eSeveral previous studies have explored XR-based navigation systems for breast tumor localization [17-19, 24]. Recently, breast DTs, created using supine MRI, were aligned to the patient using color and depth sensors, during a breast conservative surgery [18,19]. However, the visualization in these systems was limited to 2D displays, which do not capture the volumetric complexities of breast anatomy. This was attributed to the complexity of integrating 3D model registration into surgical workflows. The system described by Sharifian et al., did not acquire images in real-time; instead, the camera was rotated around the patient and three frames from the sequence were selected to register the AR contents, and then were displayed to the surgeon [18]. The slow nature of this approach limits compatibility with intraoperative workflows and scalability. Ock et al. reported accurate results on tumor localization for BCS performed on a mannequin, using a system designed for a smartphone and performing the DT alignment by detecting the 2D breast phantom\u0026rsquo;s boundary line [24].\u003c/p\u003e\n\u003cp\u003eOutside BC, XR systems have been clinically validated in procedures such as open pancreatic surgery and laparoscopic microwave ablation of hepatic hemangiomas [25, 26]. Both studies rely on manual or marker-based registration of static DTs, leading to a decline in accuracy over time due to intraoperative tissue deformation. In BCS, the lack of anatomical reference points in the breast further limits marker-based registration. Although markerless registration methods have been a focus of research, they currently underperform marker-based approaches, limiting their use in real-time clinical settings [27].\u003c/p\u003e\n\u003cp\u003eIn this work, DTs were created from supine MRI, automatically aligned to the patient in real-time using a dual-Azure Kinect 3D sensing system and visualized intraoperatively through an XR headset. Although limited by a small sample size (n=4), tumors varied considerably in volume (0.02\u0026ndash;12.74 cm\u0026sup3;), and included unifocal and multifocal cases, providing preliminary evidence across heterogeneous presentations.\u0026nbsp;All procedures were performed by an experienced breast cancer surgeon who is also a \u0026nbsp;co-author of this manuscript.\u003c/p\u003e\n\u003cp\u003eOur system accurately aligned the breast DTs with the patients\u0026rsquo; breasts with deviations in the nipple position ranging between 0 and 5 mm. In order to achieve this type of alignment, patients were positioned with the arms along the body to more closely resemble the position during the MRI supine acquisition. In one case (patient #4), however, there were visible differences between the breast DT and the patient\u0026rsquo;s breast shape in the OR, leading to the largest observed nipple deviation (5mm). This can be explained by slight changes between the patient\u0026rsquo;s position during the MRI acquisition and at the OR. Despite the overall quality of the DT alignment, the tumor projection on the skin did not necessarily correspond to the carbon tattoo marking. This was particularly evident in the most lateral tumor of patient #3, which showed a very large deviation between the vertical projection of the digital tumor and the carbon tattoo marking (\u0026gt;30mm). However, according to the surgeon, the tumor localization intraoperatively matched the one marked by the DT. This highlights the limitations of carbon tattooing, or other conventional methods such as wire-guided, as a reference standard, given their lack of standardization and non-vertical trajectory to the tumor. Accordingly, surgeon confirmation of intraoperative findings remains essential, and margin status will be required for future quantitative validation.\u003c/p\u003e\n\u003cp\u003eThe system setup time was 218 seconds (range 72-254) and tumor marking time was 49 seconds (range 32-72) during the procedures, and the surgeon reported these results as compatible with routine surgical workflow. The system was removed within seconds after use, allowing the surgery to proceed with the surgery seamlessly. Usability testing demonstrated high surgeon satisfaction (SUS score=95; range 92.5-100; rated as \u0026quot;Best Imaginable\u0026quot;), and low perceived workload, likely reflecting close co-design with clinicians and workflow prioritization. The variability in effort scores (NASA score=20; range 10-90) may reflect inconsistent cognitive demands across different tumor sizes and anatomy, requiring further investigation in larger cohorts. Progressive improvements in alignment accuracy and display latency suggest a learning curve for both surgeons and the technical team. This collaboration exemplifies a proactive step toward the emerging concept of a medical metaverse in the OR [28-30].\u003c/p\u003e\n\u003cp\u003eThe conceptual distinction between 3D and 4D surgery is critical for fully leveraging DTs. 3D static DT are reconstructed from different imaging modalities to enhance personalized surgical planning, tumor segmentation, and preoperative simulations. \u0026nbsp;4D models can extend this framework by incorporating temporal dynamics, such as respiratory motion, tissue deformation, and real-time navigation from intraoperative sensors, which are critical for XR-based navigation systems in the OR [30]. In contrast, conventional surgery planning workflows rely mainly on reports and 2D imaging analysis, imposing a high-cognitive burden on surgeons. They must keep a memory of preoperative medical images, infer tumor location on the patient\u0026apos;s breast despite differences between imaging and surgical positioning, while cognitively performing breast and tumor volumetric assessment. In this work, shadow twins are overlaid on the patient in continuous 15 fps surface scanning to track respiratory motion, supporting more accurate intraoperative decision-making. This system represents a concrete implementation of 4D surgery principles rather than just a preliminary step.\u003c/p\u003e\n\u003cp\u003eBesides providing high value to intraoperative planning, incorporating DTs into the preoperative planning workflow also enables patients to receive further validation of their concerns through treatment explanations and postoperative outcomes [31,32]. Thus, patients can be better educated and participate more proactively in the surgery decision-making process. Simultaneously, DTs can further serve as a ground-truth to validate other emerging imaging techniques, such as AI-driven 3D ultrasound [33].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFuture work should extend this proof-of-concept to a larger patient cohort, assess accuracy using margin status as the primary outcome measure, and include a larger group of surgeons including some with limited XR experience. Additionally, advances in biomechanical breast modeling and non-rigid registration are needed to accurately simulate tissue deformation and tumor displacement across different patient positions and throughout breathing motion [30,34-36]. Another current limitation is the reliance on manual segmentation by a physician with radiologist validation, which constrains scalability. Incorporating automatic segmentation of MRI images could streamline model generation, reduce operator dependency, and facilitate broader clinical adoption [37-39].\u0026nbsp;Beyond technical improvements, ethical considerations, social impact, and legal guidance remain critical considerations for safe and responsible clinical implementation [12]. Such improvements could further enhance the predictive power and clinical value of DTs in breast cancer surgery. For translation into clinical practice, compliance with the European Union Medical Device Regulation (MDR, Regulation (EU) 2017/745) will be required, including conformity assessment, clinical evaluation, and Conformit\u0026eacute; Europ\u0026eacute;enne (CE) marking to ensure safety and performance.\u003c/p\u003e\n\u003cp\u003eBy integrating DTs in XR environments, new paradigms of interaction with patient-specific data are available to surgeons and healthcare professionals overall, contributing to non-invasive and intraoperative tumor localization, with digital precision. This proof-of-concept demonstrates the feasibility of automatically aligning DTs with patients in the OR while meeting clinical workflow requirements, enabling accurate tumor localization, and enhanced surgical performance.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eEthical approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients were enrolled in the study following study protocol presentation and signed informed consent. The study was performed in compliance with relevant laws and institutional guidelines and was approved by the ethics committee of Champalimaud Foundation on March 2022 with the identification \u003cem\u003eBreast 4.0 v 3.0\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e(no applicable approval number).\u003c/p\u003e\n\u003cp\u003ePatient-specific Digital Twins for Breast Cancer Surgery\u003c/p\u003e\n\u003cp\u003eFour patients were selected and consented for acquisition of contrast-enhanced MRI images. Contrast-enhanced MRI images were taken after the administration of a contrast agent to improve lesion visualization. An additional scan in the supine position was acquired after the standard prone sequence to build the DTs with the patient in a position similar to the one in the OR (with the arms along the body).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAll scans were performed on a 3T MR system, using a dedicated 16-channel breast coil (for the prone acquisition) and for the supine acquisition a 32-channel torso coil was used together with a position device that allows the usage of the coil without distorting the anatomy of the patient. The supine acquisition was a 3D T1-weighted Dixon scan of the thorax with breathing compensation and the following scan parameters: TE=1.57 ms; TR=5.3 ms; FOV=450x450 mm^2, slices=250; in-plane resolution=1.2x1.2 mm^2; slice thickness=1.5 mm; and a total scan time=05:52 s.\u003c/p\u003e\n\u003cp\u003eSkin, fibroglandular tissue, and tumor segmentations were performed on the supine MRI images using 3D Slicer by a physician and validated by an experienced radiologist (Figure 2). The patient\u0026rsquo;s image data was anonymized, stored in a research database, and used to create a static DT of the patient. An additional scan of the patient\u0026rsquo;s skin surface was acquired using the Go!Scan 50 handheld scanning device, in supine position, for validation and prior testing of the registration algorithm before surgery.\u003c/p\u003e\n\u003cp\u003ePost segmentation, the 3D model was exported and modelled using the Open3D library in Python. Since the segmentation output contained the whole volume of the breast (Figure 4a), the model\u0026rsquo;s surface was first extracted. By default, 10,000 rays were cast from the outside to the center of the mesh, with each collision resulting in a vertice on the final mesh (Figure 4b). Rays near the poles were removed to avoid intersections with points inside the breast volume. Statistical outliers were removed by calculating the mean distance between each point and its neighboring points. The global mean and standard deviation of these average distances were computed with points with an average distance greater than \u003cem\u003emean + (std_ratio \u0026times; standard deviation)\u003c/em\u003e being classified as outliers and excluded from further analysis (Figure 4c). The surface was clipped along the sagittal axis to remove areas on the side and back.\u003c/p\u003e\n\u003cp\u003eOptimized XR interface\u003c/p\u003e\n\u003cp\u003eThe DT was uploaded to the XR application and visualized through the ML. Co-design sessions were conducted with surgeons in the OR to inform the XR interface design. Early concepts were implemented in Unity 3D with MRTK3 and iteratively based on surgeons\u0026rsquo; feedback (Figure 5a-b). This process focused on optimizing tumor visualization for accurate marking, leading to the adoption of visual cues, such as a skin grid and tumor projection onto the surface (Figure 6 and Supplementary video).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe interface menu was spatially anchored to the DT, constrained to move within a viewing frustum defined relative to the DT\u0026rsquo;s coordinate system, oriented toward the surgeon (Figure 5c and Supplementary video). Positioned at an approximate viewing distance of 60 cm from the surgeon, the interface supported head-gaze targeting and pinch-based selection (Figure 5c-d). Throughout iterative refinement, surgeon feedback and prototype development converged towards a high-fidelity interface suitable for use in the OR.\u003c/p\u003e\n\u003cp\u003eSurgeons emphasized the importance of visualizing the tumor projection on the skin, given the lack of standardization in conventional carbon tattooing. Tumor marking was defined as a vertical vector from the tumor center to the skin surface, parallel to gravity\u0026rsquo;s vector (Figure 6a-b). The projection was considered helpful during tumor marking. Additionally, the ML\u0026rsquo;s near clipping boundary (~37 cm) was intentionally used as a visual aid to better understand tumor depth and position (Figure 6c-d and supplementary video). This provided an easy way for the surgeon to switch the visualization of the digital content on or off by moving further or closer to the patient.\u003c/p\u003e\n\u003cp\u003e3D sensing system for DT to patient-alignment\u003c/p\u003e\n\u003cp\u003eTo align the patient\u0026rsquo;s breast DT to the patient in real-time in the OR, a 3D surface scan system composed of two Azure Kinect cameras (Figure 7a) was developed and validated with patient data [40]. The system was designed to be attached to the operating table and incorporated placeholders for AprilTag markers. These markers establish a common coordinate system between the Azure Kinects and the ML headset. Upon patient detection by the 3D surface scan system, the DT is automatically registered to the patient using an interactive closest point (ICP) algorithm and displayed in the XR interface (Figure 7b).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe surface scanning system continually updates the DT position (15 fps) displayed to the surgeon through the ML, to keep up with the patient\u0026rsquo;s breathing motion. In this configuration, the DT operates not as a static model but as a dynamic shadow twin, integrating real-time data to maintain accurate alignment with the patient [11]. Then, the surgeon selects the option to freeze the DT position at the instant of the marking task, to ensure its stability during those seconds. All procedures were performed by a breast surgeon with over 15 years of experience in breast-conserving surgery, who is a co-author of this study. The surgeon had prior experience with the XR system before clinical implementation.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe breast supine MRIs, tissue segmentations, and DTs for the four patients are available in this study repository, LINK_TO_REPOSITORY.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe underlying code for the XR system and DTs alignment in the OR is not publicly available for proprietary reasons.\u003c/p\u003e\n\u003cp\u003e\u003cb\u003eAcknowledgements\u003c/b\u003e\u003c/h3\u003e\n\u003cp\u003eThis research is part of the Health for Portugal (HfP), funded by Agendas Mobilizadoras para a Inova\u0026ccedil;\u0026atilde;o Empresarial - Plano de Recupera\u0026ccedil;\u0026atilde;o e Resili\u0026ecirc;ncia (PRR) Portugu\u0026ecirc;s (no applicable grant number). The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. The authors would like to thank the Breast Unit at Champalimaud Foundation for their cooperation in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRT, JS, TM, and PG conceptualized and designed the study. RT acquired the patients\u0026rsquo; surface scan prior to surgery. AC and MC performed the breast segmentations. NL, YF, and JS managed the medical imaging processing pipeline. BV, AL, and TM designed and implemented the DTs production pipeline and the XR system for the OR. RT, DSL, and TM led the user interface design. RT and TM analyzed and interpreted the data. RT, TM, and PG prepared the original draft. All authors reviewed and edited the manuscript and read and approved the final version.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthors JS, TM, and PG have a provisional patent application - METHODS AND SYSTEMS FOR PRECISION-GUIDED SURGERY - pending to Champalimaud Foundation but declare no other competing interests. All other authors declare no financial or non-financial competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eThe International Agency for Research on Cancer (IARC). (n.d.). \u003cem\u003eGlobal Cancer Observatory\u003c/em\u003e. https://gco.iarc.who.int/en. Last accessed 2024/02/08.\u003c/li\u003e\n\u003cli\u003eCardoso, F., Kyriakides, S., Ohno, S., Penault-Llorca, F., Poortmans, P., Rubio, I. T., ... \u0026amp; Senkus, E. (2019). Early breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. \u003cem\u003eAnnals of oncology\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(8), 1194-1220.\u003c/li\u003e\n\u003cli\u003eYu, C. X. (2023). Radiotherapy of early‐stage breast cancer. \u003cem\u003ePrecision Radiation Oncology\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(1), 67-79.\u003c/li\u003e\n\u003cli\u003eCorsi F, Sorrentino L, Bossi D, Sartani A, Foschi D. Preoperative localization and surgical margins in conservative breast surgery. Int J Surg Oncol 2013;2013: 793819.\u003c/li\u003e\n\u003cli\u003eRyan, J. F., Ghosh, S., Rajaee, A. N., Olson, D., Lesniak, D. M., Peiris, L. J., \u0026amp; Alberta Breast Surgical Oncology Research Group (2024). Re-excision rates after breast-conserving surgery for invasive breast cancer: an Albertan perspective. Canadian journal of surgery. Journal canadien de chirurgie, 67(6), E363\u0026ndash;E369. https://doi.org/10.1503/cjs.004723\u003c/li\u003e\n\u003cli\u003eW. L. Richey, J. S. Heiselman, M. J. Ringel, I. M. Meszoely and M. I. Miga, \u0026quot;Computational Imaging to Compensate for Soft-Tissue Deformations in Image-Guided Breast Conserving Surgery,\u0026quot; in \u003cem\u003eIEEE Transactions on Biomedical Engineering\u003c/em\u003e, vol. 69, no. 12, pp. 3760-3771, Dec. 2022, doi: 10.1109/TBME.2022.3177044.\u003c/li\u003e\n\u003cli\u003eBessa, S., Gouveia, P. F., Carvalho, P. H., Rodrigues, C., Silva, N. L., Cardoso, F., ... \u0026amp; Cardoso, M. J. (2020). 3D digital breast cancer models with multimodal fusion algorithms. \u003cem\u003eThe Breast\u003c/em\u003e, \u003cem\u003e49\u003c/em\u003e, 281-290.\u003c/li\u003e\n\u003cli\u003eGarzotto, F., Comoretto, R. I., Michieletto, S., Franzoso, G., Mele, M. L., Gregori, D., ... \u0026amp; Saibene, T. (2021). Preoperative non-palpable breast lesion localization, innovative techniques and clinical outcomes in surgical practice: A systematic review and meta-analysis. \u003cem\u003eThe Breast\u003c/em\u003e, \u003cem\u003e58\u003c/em\u003e, 93-105.\u003c/li\u003e\n\u003cli\u003eBarth, R. J., Jr, Krishnaswamy, V., Paulsen, K. D., Rooney, T. B., Wells, W. A., Angeles, C. V., Zuurbier, R. A., Rosenkranz, K., Poplack, S., \u0026amp; Tosteson, T. D. (2019). A Randomized Prospective Trial of Supine MRI-Guided Versus Wire-Localized Lumpectomy for Breast Cancer. \u003cem\u003eAnnals of surgical oncology\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(10),3099\u0026ndash;3108. https://doi.org/10.1245/s10434-019-07531-4\u003c/li\u003e\n\u003cli\u003eFausto, A., Fanizzi, A., Volterrani, L., Mazzei, F. G., Calabrese, C., Casella, D., Marcasciano, M., Massafra, R., La Forgia, D., \u0026amp; Mazzei, M. A. (2020). Feasibility, Image Quality and Clinical Evaluation of Contrast-Enhanced Breast MRI Performed in a Supine Position Compared to the Standard Prone Position. \u003cem\u003eCancers\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(9), 2364. https://doi.org/10.3390/cancers12092364\u003c/li\u003e\n\u003cli\u003eZhang, K., Zhou, H. Y., Baptista-Hon, D. T., Gao, Y., Liu, X., Oermann, E., ... \u0026amp; Wu, J. (2024). Concepts and applications of digital twins in healthcare and medicine. \u003cem\u003ePatterns\u003c/em\u003e, \u003cem\u003e5\u003c/em\u003e(8).\u003c/li\u003e\n\u003cli\u003eMekki, Y.M., Luijten, G., Hagert, E. \u003cem\u003eet al.\u003c/em\u003e Digital twins for the era of personalized surgery. \u003cem\u003enpj Digit. Med.\u003c/em\u003e 8, 283 (2025). https://doi.org/10.1038/s41746-025-01575-5.\u003c/li\u003e\n\u003cli\u003eShi, Y., Deng, X., Tong, Y., Li, R., Zhang, Y., Ren, L., \u0026amp; Si, W. (2022). Synergistic digital twin and holographic augmented-reality-guided percutaneous puncture of respiratory liver tumor. \u003cem\u003eIEEE Transactions on Human-Machine Systems\u003c/em\u003e, \u003cem\u003e52\u003c/em\u003e(6), 1364-1374.\u003c/li\u003e\n\u003cli\u003eKolekar, A., Shalgar, S., \u0026amp; Malawade, I. (2023, September). Beyond reality: A study of integrating digital twins. In \u003cem\u003eJournal of Physics: Conference Series\u003c/em\u003e (Vol. 2601, No. 1, p. 012030). IOP Publishing.\u003c/li\u003e\n\u003cli\u003eKatsoulakis, E., Wang, Q., Wu, H. \u003cem\u003eet al.\u003c/em\u003e Digital twins for health: a scoping review. \u003cem\u003enpj Digit. Med.\u003c/em\u003e 7, 77 (2024). https://doi.org/10.1038/s41746-024-01073-0.\u003c/li\u003e\n\u003cli\u003eAsciak, L., Kyeremeh, J., Luo, X. \u003cem\u003eet al.\u003c/em\u003e Digital twin assisted surgery, concept, opportunities, and challenges. \u003cem\u003enpj Digit. Med.\u003c/em\u003e 8, 32 (2025). https://doi.org/10.1038/s41746-024-01413-0.\u003c/li\u003e\n\u003cli\u003eGouveia, P. F., Costa, J., Morgado, P., Kates, R., Pinto, D., Mavioso, C., ... \u0026amp; Cardoso, M. J. (2021). Breast cancer surgery with augmented reality. \u003cem\u003eThe Breast\u003c/em\u003e, \u003cem\u003e56\u003c/em\u003e, 14-17.\u003c/li\u003e\n\u003cli\u003eSharifian, R., Madad-Zadeh, S., Bourdel, N., Giro, A., Marraoui, W., Pomel, C., \u0026amp; Bartoli, A. (2024, October). Guiding Breast Conservative Surgery by Augmented Reality from Preoperative MRI: Initial System Design and Retrospective Trials. In \u003cem\u003eDeep Breast Workshop on AI and Imaging for Diagnostic and Treatment Challenges in Breast Care\u003c/em\u003e (pp. 212-220). Cham: Springer Nature Switzerland.\u003c/li\u003e\n\u003cli\u003eLee, M., Woo, J., Peak, S. H., Kim, H. G., Lim, W. S., Chung, J., ... \u0026amp; Lee, J. W. (2024). An exploratory clinical trial of preoperative non-invasive localization before breast-conserving surgery using augmented reality technology. \u003cem\u003eBreast Cancer Research and Treatment\u003c/em\u003e, \u003cem\u003e206\u003c/em\u003e(1), 31-44.\u003c/li\u003e\n\u003cli\u003eHart, S. G., \u0026amp; Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In Advances in psychology (Vol. 52, pp. 139-183). North-Holland.\u003c/li\u003e\n\u003cli\u003eByers, J. C., Bittner, A. C., \u0026amp; Hill, S. G. (1989). Traditional and raw task load index (TLX) correlations: Are paired comparisons necessary. Advances in industrial ergonomics and safety, 1, 481-485.\u003c/li\u003e\n\u003cli\u003eBrooke, J. (1996). SUS-A quick and dirty usability scale. Usability evaluation in industry, 189(194), 4-7.\u003c/li\u003e\n\u003cli\u003eBangor, A., Kortum, P., \u0026amp; Miller, J. (2009). Determining what individual SUS scores mean: Adding an adjective rating scale. Journal of usability studies, 4(3), 114-123.\u003c/li\u003e\n\u003cli\u003eOck, J., Moon, S., Kim, M., Ko, B. S., \u0026amp; Kim, N. (2024). Evaluation of the accuracy of an augmented reality-based tumor-targeting guide for breast-conserving surgery. \u003cem\u003eComputer Methods and Programs in Biomedicine\u003c/em\u003e, \u003cem\u003e245\u003c/em\u003e, 108002.\u003c/li\u003e\n\u003cli\u003eLan, T., Liu, S., Dai, Y., Luo, J., Han, J., \u0026amp; Jin, Y. (2025). Real-time image fusion and Apple Vision Pro in laparoscopic microwave ablation of hepatic hemangiomas. \u003cem\u003eNPJ Precision Oncology\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1), 79.\u003c/li\u003e\n\u003cli\u003eJavaheri, H., Ghamarnejad, O., Lukowicz, P., Stavrou, G. A., \u0026amp; Karolus, J. (2025, April). From Concept to Clinic: Multidisciplinary Design, Development, and Clinical Validation of Augmented Reality-Assisted Open Pancreatic Surgery. In \u003cem\u003eProceedings of the 2025 CHI Conference on Human Factors in Computing Systems\u003c/em\u003e (pp. 1-24).\u003c/li\u003e\n\u003cli\u003eM. Tu, H. Jung, A. Moghadam, J. Raythatha, J. Hsu and J. Kim, \u0026quot;Exploring the Performance of Geometry-Based Markerless Registration in a Simulated Surgical Environment: A Comparative Study of Registration Algorithms in Medical Augmented Reality,\u0026quot; \u003cem\u003e2023 45th Annual International Conference of the IEEE Engineering in Medicine \u0026amp; Biology Society (EMBC)\u003c/em\u003e, Sydney, Australia, 2023, pp. 1-4, doi: 10.1109/EMBC40787.2023.10341197.\u003c/li\u003e\n\u003cli\u003eNavab, N., Martin-Gomez, A., Seibold, M., Sommersperger, M., Song, T., Winkler, A., ... \u0026amp; Eck, U. (2022). Medical augmented reality: definition, principle components, domain modeling, and design-development-validation process. \u003cem\u003eJournal of Imaging\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1), 4.\u003c/li\u003e\n\u003cli\u003eGouveia, P. F., Luna, R., Fontes, F., Pinto, D., Mavioso, C., Anacleto, J., ... \u0026amp; Cardoso, M. J. (2023). Augmented reality in breast surgery education. \u003cem\u003eBreast Care\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(3), 182-186.\u003c/li\u003e\n\u003cli\u003eRiedel, P., Riesner, M., Wendt, K., \u0026amp; A\u0026szlig;mann, U. (2022, May). Data-driven digital twins in surgery utilizing augmented reality and machine learning. In \u003cem\u003e2022 IEEE International Conference on Communications Workshops (ICC Workshops)\u003c/em\u003e (pp. 580-585). IEEE.\u003c/li\u003e\n\u003cli\u003eKaidar-Person, O., Antunes, M., Cardoso, J. S., Ciani, O., Cruz, H., Di Micco, R., ... \u0026amp; CINDERELLA Consortium. (2023). Evaluating the ability of an artificial-intelligence cloud-based platform designed to provide information prior to locoregional therapy for breast cancer in improving patient\u0026rsquo;s satisfaction with therapy: The CINDERELLA trial. \u003cem\u003ePLoS One\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(8), e0289365.\u003c/li\u003e\n\u003cli\u003eBonci, E. A., Kaidar-Person, O., Antunes, M., Ciani, O., Cruz, H., Di Micco, R., ... \u0026amp; Cardoso, M. J. (2024). CINDERELLA Trial: validation of an artificial-intelligence cloud-based platform to improve the shared decision-making process and outcomes in breast cancer patients proposed for locoregional treatment. \u003cem\u003eEuropean Journal of Surgical Oncology\u003c/em\u003e, \u003cem\u003e50\u003c/em\u003e(2).\u003c/li\u003e\n\u003cli\u003eKwon, H., Oh, S. H., Kim, M. G., Kim, Y., Jung, G., Lee, H. J., ... \u0026amp; Bae, H. M. (2024). Enhancing Breast Cancer Detection through Advanced AI-Driven Ultrasound Technology: A Comprehensive Evaluation of Vis-BUS. \u003cem\u003eDiagnostics\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(17), 1867.\u003c/li\u003e\n\u003cli\u003eCarolina, L., Timoteo, R., Laborde, A., Vaz, B., Gouveia, P., Marques, T. (2025). Volumetric Correction for Aligning Breast Digital Twins with Surface Scan. \u003cem\u003eProceedings of the 14th Conference on New Technologies for Computer/Robot Assisted Surgery.\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eTeixeira, A. M., \u0026amp; Martins, P. (2023). A review of bioengineering techniques applied to breast tissue: Mechanical properties, tissue engineering and finite element analysis. \u003cem\u003eFrontiers in bioengineering and biotechnology\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e, 1161815. https://doi.org/10.3389/fbioe.2023.1161815\u003c/li\u003e\n\u003cli\u003eEiben, B., Vavourakis, V., Hipwell, J. H., Kabus, S., Buelow, T., Lorenz, C., Mertzanidou, T., Reis, S., Williams, N. R., Keshtgar, M., \u0026amp; Hawkes, D. J. (2016). Symmetric Biomechanically Guided Prone-to-Supine Breast Image Registration. Annals of biomedical engineering, 44(1), 154\u0026ndash;173. https://doi.org/10.1007/s10439-015-1496-z \u003c/li\u003e\n\u003cli\u003eForghani, Y., Tim\u0026oacute;teo, R., Marques, T., Lou\u0026ccedil;\u0026atilde;o, N., Cardoso, M. J., Cardoso, F., ... \u0026amp; Santinha, J. (2025). Comparative analysis of nnU-Net and Auto3Dseg for fat and fibroglandular tissue segmentation in MRI. \u003cem\u003eJournal of Medical Imaging\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(2), 024005-024005.\u003c/li\u003e\n\u003cli\u003eLew, C.O., Harouni, M., Kirksey, E.R. \u003cem\u003eet al.\u003c/em\u003e A publicly available deep learning model and dataset for segmentation of breast, fibroglandular tissue, and vessels in breast MRI. \u003cem\u003eSci Rep\u003c/em\u003e 14, 5383 (2024). https://doi.org/10.1038/s41598-024-54048-2\u003c/li\u003e\n\u003cli\u003eNarimani, S., Roth Hoff, S., D\u0026aelig;hli Kurz, K. \u003cem\u003eet al.\u003c/em\u003e Comparative analysis of deep learning architectures for breast region segmentation with a novel breast boundary proposal. \u003cem\u003eSci Rep\u003c/em\u003e 15, 8806 (2025). https://doi.org/10.1038/s41598-025-92863-3\u003c/li\u003e\n\u003cli\u003eTimoteo, R., Laborde, A., Forghani, Y., Lopes, D.S., Gouveia, P., Marques, T. (2025). Automatic Surface Scan System for Breast Cancer Surgery. \u003cem\u003eProceedings of the 14th Conference on New Technologies for Computer/Robot Assisted Surgery\u003c/em\u003e.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8907845/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8907845/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Breast-conserving surgery followed by adjuvant radiotherapy is the standard treatment for breast cancer, yet inaccurate tumor localization may lead to positive margins and re-excision. Current localization techniques are invasive and often imprecise. Digital twins (DT), patient-specific digital replicas, can offer a non-invasive approach for tumor localization. In this proof-of-concept study, breast DTs were created from supine magnetic resonance imaging and visualized through an extended reality (XR) headset in four breast conservative surgeries. DTs were automatically registered to the patient in real time using a 3D sensing system. The accuracy of the alignment deviations between the nipple and its position in the DT ranged 0-5 mm. The surgeon reported that the tumor XR visualization was consistent with intraoperative findings. Setup and marking times were clinically acceptable, and preliminary usability assessments indicated high perceived usability and low workload. These results support the feasibility of integrating DTs into breast cancer surgical workflows.","manuscriptTitle":"Digital twins for breast cancer surgery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-27 07:35:20","doi":"10.21203/rs.3.rs-8907845/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8629156b-4c25-457d-92b8-fad85eda5b21","owner":[],"postedDate":"February 27th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63325513,"name":"Biological sciences/Cancer"},{"id":63325514,"name":"Health sciences/Health care"},{"id":63325515,"name":"Health sciences/Medical research"},{"id":63325516,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2026-02-27T07:35:20+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-27 07:35:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8907845","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8907845","identity":"rs-8907845","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}