Consistent Point Matching

Abstract This study demonstrates that incorporating a consistency heuristic into the point-matching algorithm improves robustness in thetask of matching anatomical locations in pairs of medical images. We validated our approach on diverse longitudinal internaland public datasets spanning CT and MRI modalities. Notably, it surpasses state-of-the-art results on the Deep Lesion Trackingdataset. Additionally, we have shown that the method effectively addresses landmark localization. The algorithm operatesefficiently on standard CPU hardware, allowing configurable trade-offs between speed and robustness. It enables automatednavigation in medical images while eliminating the need for machine learning pipelines.