title:: Probabilistic Sparse Matching for Robust 3D/3D Fusion in Minimally Invasive Surgery
publisher:: IEEE
people:: Dominik Neumann, Sasa Grbic, Matthias John, Nassir Navab, Joachim Hornegger, Razvan Ionasec
organization:: Siemens Corporate Research, Siemens Healthineers, Technical University of Munich, Friedrich-Alexander-Universität Erlangen-Nürnberg
domain:: Medical Imaging, Machine Learning, Computer Vision, Image-Guided Intervention
link:: https://doi.org/10.1109/tmi.2014.2343936
citations:: 11 (OpenAlex) · 11 (Semantic Scholar) — as of 14 October 2026
Article record — Europe PMC.
Summary
Transcatheter procedures give the operator no direct view of the anatomy, so intraoperative C-arm CT and fluoroscopy carry the guidance - with limited soft tissue quality, and a contrast agent requirement that is harmful to the patient and complicates the acquisition protocol.
This paper fuses high-quality preoperative CT with non-gated, non-contrast intraoperative C-arm CT using a probabilistic sparse matching approach built on machine learning and numerical optimisation. The patient-specific model extracted preoperatively is mapped into the intraoperative imaging environment, so the guidance inherits the preoperative image quality without needing contrast during the procedure.
Data Points
- Fuses preoperative CT with non-gated, non-contrast intraoperative C-arm CT - removing the intraoperative contrast requirement
- Model-based rather than image-to-image: the preoperative patient-specific model is what gets mapped
- Validated on 95 clinical datasets; average execution time 1.56 s
- Accuracy of 5.48 mm between anchor anatomy in both images, within expert user confidence intervals
- Outperforms both an open-source state-of-the-art registration library and a quasi-global knowledge-driven multi-modal fusion method on accuracy and robustness
Reference
IEEE Transactions on Medical Imaging 34(1):49-60, 2014. doi:10.1109/tmi.2014.2343936 · PubMed 25095250
11 citations (OpenAlex and Semantic Scholar agree, October 2026)
Part of the complete record in Scientific Publications - Peer-Reviewed Research Record.
