Academic Library

Bruneau, R., Brument, B., Quéau, Y., Mélou, J., Lauze, F. B., Durou, J. & Calvet, L.

International Journal of Computer Vision (2026)

In this study a versatile reconstruction framework integrates multi-view normal and reflectance information directly into radiance-based surface reconstruction pipelines. By jointly re-parameterising surface normals and reflectance as radiance vectors under varying illumination, the approach can be seamlessly combined with both traditional multi-view stereo and neural volume rendering methods. Extensive benchmarking shows state-of-the-art performance on multiple MVPS datasets, with particularly strong results for fine-detail recovery and challenging visibility conditions.

Read more

Meyer, J., Giraud, F., Wüthrich, J., Pollefeys, M., Fürnstahl, P. & Calvet, L.

Sensors (2026)

RocSync delivers a millisecond-accurate approach to synchronizing heterogeneous RGB and IR camera systems. This study uses a custom LED clock to encode exposure timing, achieving a residual error of 1.34 ms and outperforming light-, audio-, and timecode-based methods. The approach enhances multi-view pose estimation and 3D reconstruction and was validated in large-scale surgical recordings with more than 25 cameras.

Read more

Wu, L., Seibold, M., Cavalcanti, N., Loggia, G., Reissner, L., Sigirst, B., Hein., J., Calvet, L., Viehöfer, A. & Fürnstahl, P.

Computerized Medical Imaging and Graphics (2026)

UltraBoneUDF introduces a self-supervised framework for reconstructing open bone surfaces from ultrasound using neural unsigned distance functions. By fusing ultrasound-specific features and optimizing local surface geometry, the method overcomes limitations of partial bone capture in B-mode ultrasound. Extensive evaluation on multiple datasets shows substantial improvements over existing techniques, highlighting the potential of ultrasound-based 3D reconstruction for safer and more accessible orthopedic surgery.

Read more

Wu, L., Seibold, M., Cavalcanti, N., Hein, J., Gerth, T., Lekar, R., Hoch, A., Vlachopoulos, L., Grabner, H., Zingg, P., Farshad, M. & Fürnstahl, P.

Computers in Biology and Medicine (2025)

This study presents an AR-based simulator for Total Hip Arthroplasty (THA), reducing reliance on instructors. The simulator uses AR guidance and automated evaluation for self-paced training. Feasibility results showed mean deviation errors below 3 mm and 3 degrees, with no significant difference from manual assessment, demonstrating the potential of AR simulators to enhance orthopedic training efficiency and safety.

Read more

Most, E., Hein, J., Giraud, F., Cavalcanti, N., Zingg, L., Brument, B., Louman, N., Carrillo, F., Fürnstahl, P. & Calvet, L.

International Journal of Computer Assisted Radiology and Surgery (2025)

This study presents an approach for generating realistic ex vivo datasets with 3D ground truth, tailored to optical image-based reconstruction and feature matching in open orthopedic surgery. By addressing the current lack of suitable data, it supports the development of applications such as marker-less surgical navigation and surgical digitization.

Read more

Bahari Malayeri, A., Seibold, M., Cavalcanti, N., Hein, J., Jecklin, S., Vlachopoulos, L., Fucentese, S., Hodel, S. & Fürnstahl, P.

Computer Assisted Surgery (2025)

Surgical phase recognition in arthroscopy is advanced through the introduction of the first dedicated dataset (ACL27) and a transformer-based model. By leveraging spatio-temporal features, the method addresses challenges such as limited field of view, occlusions, and visual distortions, while a surgical progress index (sPi) enables procedure progression estimation. Results on ACL27 and Cholec80 confirm strong performance and generalizability, underlining the potential for improved training, real-time assistance, and efficiency in orthopedic surgery.

Read more

Flückiger, T., Hein, J., Fischer, V., Fürnstahl, P. & Calvet, L.

International Journal of Computer Assisted Radiology and Surgery (2025)

A novel automated calibration method is introduced for multi-camera systems in 3D surgical scene reconstruction, eliminating the need for manual intervention or specialized expertise. Using a ceiling-mounted projector to display multi-scale markers, the approach achieves accuracy comparable to manual methods while offering greater robustness under varying zoom levels. Validation on synthetic and real OR data highlights its potential to enable fully automated 3D-SSR.

Read more

Flepp, R., Nissen, L., Sigrist, B., Nieuwland, A., Cavalcanti, N., Fürnstahl, P., Dreher, T. & Calvet, L.

International Journal of Computer Assisted Radiology and Surgery (2025)

A novel multi-view X-ray/CT registration method is presented to improve intraoperative bone registration in orthopedic surgery. Unlike existing techniques, it is designed to achieve high accuracy and robustness without relying on manual key-point annotations, addressing key limitations in current surgical navigation approaches.

Read more

Farshad, M., Schader, J., Stauffer, A., Zipser, C., Kheram, N., Spirig, J., Fasser, M., Widmer, J. & Hagel, V.

Neurospine (2025)

This study investigated the effect of irrigation fluid pressure during interlaminar endoscopic lumbar discectomy (IELD) on spinal intra-/epidural and intracranial pressures. Experiments on human cadavers showed a linear correlation between irrigation settings and pressure increases across spinal and intracranial levels. While pressures remained within safe ranges under normal conditions, outflow occlusion and the presence of a dural tear led to significantly higher values, highlighting important safety considerations for endoscopic spine surgery.

Read more

Jecklin, S., Massalimova, A., Zha, R., Calvet, L., Laux, C., Farshad, M. & Fürnstahl, P.

Scientific Reports (2025)

This work extends Gaussian splatting to enable anatomically consistent 3D spinal reconstruction from sparse and arbitrarily posed intraoperative X-rays. By introducing an anatomy-guided radiographic standardization step using style transfer, the method improves visual consistency and reconstruction quality while requiring no pretraining, making it inherently adaptable to new patients and anatomies. Evaluation on an ex-vivo dataset and expert surgical assessment confirm its clinical utility for navigation, demonstrating the feasibility of instance-based volumetric reconstruction under real surgical conditions.

Read more

Huber, M., Cavalcanti, N., Davoodi, A., Li, R., Mower, C. E., Carrillo, F., Laux, C. J., Teyssere, F., Chandanson, T., Harlé, A., Saghbiny, E., Farshad, M., Morel, G., Vander Poorten, E., Fürnstahl, P., Ourselin, S., Bergeles, C. & Vercauteren, T.

arXiv (2025)

Limited spatial awareness remains a key barrier to safer and more autonomous surgical robotics. This work introduces a marker-free proprioception approach that enables precise localization of fully draped surgical robots using only lightweight stereo-RGB cameras and transformer-based deep learning models. Trained on the largest multi-centre robotic surgery dataset to date, the method tracks the entire robot and surrounding scene to provide a holistic, occlusion-robust spatial representation. Experiments demonstrate improved tracking visibility compared to existing systems and support multi-robot localization and context-aware control, highlighting the potential of marker-free robot perception for intelligent robotic surgery.

Read more

Wu, L., Seibold, M., Cavalcanti, N., Ao, Y., Flepp, R., Massalimova, A., Calvet, L. & Fürnstahl, P.

arXiv (2025)

A self-supervised, modality-agnostic approach is presented to align preoperative bone models with intraoperative surface data for orthopedic navigation. NeuralBoneReg represents bone anatomy as a neural unsigned distance field and aligns intraoperative point clouds through learned global initialization and local refinement, without relying on inter-subject training data. Experiments across CT–ultrasound and CT–RGB-D datasets of the tibia, fibula, hip, pelvis, and spine show consistent performance equal to or exceeding existing methods, demonstrating strong generalizability and reliability for cross-modal bone surface registration in computer- and robot-assisted orthopedic surgery (CAOS).

Read more

Hein, J., Vlachopoulos, L., Olthof, M. G., Loggia, G., Sigrist, B., Fürnstahl, P. & Seibold, M.

arXiv (2025)

A novel multimodal framework integrates 3D acoustic information with visual data to enhance surgical scene understanding. The study combines spatial sound localization with dynamic RGB-D point clouds and uses a transformer-based model to detect tool–tissue interactions, enabling temporally and spatially aware 4D scene representations. Evaluation in a realistic operating room setup demonstrates accurate 3D sound localization and robust multimodal fusion, highlighting the potential of audio-visual modeling for future intelligent and autonomous surgical systems.

Read more

Jokeit, M., Cornaz, F., Schader, J.F., Harshbarger, C.L., Calek, A.K., Jecklin, S., Snedeker, J.G., Farshad, M. & Widmer, J.

European Spine Journal (2025)

In full spinal flexion, paraspinal muscles are largely inactive, suggesting that passive structures play a key role in stability. To investigate this, a stepwise resection study was performed on cadaveric torsos, where posterior lumbar elements were gradually removed under controlled flexion. Load measurements after each step allowed assessment of the stabilizing contribution of individual structures and the potential impact of their surgical resection.

Read more

Yavuz, A., Gultekin, C., Wang, X., Pollefeys, P., Seibold, M., Fürnstahl, P., Carrillo, F. & Kaufmann, M.

9th International Conference on Intelligent Computing and Virtual & Augmented Reality Simulations (2025)

The study explores egocentric video data from head-worn cameras as a foundation for building computational models with a holistic understanding of surgical scenes and interactions. It analyzes data acquisition and processing pipelines, assesses image annotation quality, and evaluates state-of-the-art segmentation as well as 3D hand and body pose estimation models. The results highlight the challenges of real clinical environments while demonstrating the potential of egocentric datasets for advancing intelligent surgical assistance systems.

Read more

Jecklin, S., Shen, Y., Gout, A, Suter, D., Calvet, L., Zingg, L., Straub, J., Cavalcanti, N., Farshad, M., Fürnstahl, P., Esfandiari, H.

Medical Image Analysis (2024)

Generating 3D spinal models from only a few intraoperative X-ray images can eliminate the need for conventional navigation systems. By bridging synthetic and real data with advanced deep learning, this approach moves toward accurate, real-time 3D reconstruction with promising implications for surgical planning, navigation, and robotics.

Read more

Seibold, M., Bahari Malayeri, A., Fürnstahl, P.

International Conference on Medical Image Computing and Computer Assisted Intervention – MICCAI (2024)

In this study, the novel concept of Sound Source Localization (SSL) is introduced for surgery, enabling not only the detection but also the spatial mapping of acoustic activity in the operating field. Proof-of-concept experiments with an acoustic camera demonstrated the feasibility of using sound for object and keypoint detection in surgical tasks. These findings highlight SSL’s potential to advance surgical scene understanding and pave the way for multimodal sensing in future operating rooms.

Read more

Luchmann, D., Jecklin, S., Cavalcanti, N.A., Laux, C., Massalimova, A., Esfandiari, H., Farshad, M., Fürnstahl, P.

BMC Musculoskeletal Disorders (2024)

With the increasing number of spinal instrumentation surgeries, 3D navigation seeks to improve implant accuracy. Yet, adoption has been limited by radiation exposure, costs, and workflow disruption. In this study, a new approach generates 3D spinal models from only a few intraoperative X-ray images, moving toward accurate, real-time reconstruction with potential benefits for surgical planning, navigation, and robotics.

Read more

Cavalcanti, N., Carrillo, F., Li, R., van Assche, K., Davoodi, A., Tummers, M., Huber, M., Teyssere, F., Perez Velásquez, J., Massalimova, A., Laux, C., Sutter, R., Farshad, M., Borghesan, G., Denis, K., Morel, G., Chandanson, T., Vercauteren, T., Vander Poorten, E. & Fürnstahl P.

MICCAI Conference on Medical Image Computing and Computer Assisted Invervention (2024)

This study investigates the FAROS robotic system's ability to autonomously execute pedicle screw trajectories using 3D reconstructed ultrasound navigation. Results show that the system can achieve clinically acceptable accuracy, highlighting its potential for future robotic applications in surgery.

Read more

Hein, J., Grunder, J., Calvet, L., Giraud, F., Cavalcanti, N., Carrillo, F., and Fürnstahl, P.

IEEE International Symposium on Mixed and Augmented Reality Adjunct (2024)

SurgTwinVR introduces a Virtual Reality application that immerses users in a Surgical Digital Twin, creating a high-fidelity replica of the entire operating room. Unlike existing VR or AR systems, it dynamically reproduces the surgical scene while optimized for real-time rendering. This approach highlights the potential of VR-based digital twins for surgical training and education.

Read more

Seibold, M., Spirig, J., Esfandiari, H., Farshad, M., Fürnstahl, P.

Journal of Imaging (2023)

Medical Augmented Reality (AR) is beginning to make its way into patient care after decades of research. This concept paper details the translational pathway from clinical need to validation and approval, highlighting why many innovations remain prototypes. It also introduces the first guideline designed to support the successful transfer of medical AR research into clinical practice.