Tags:Deep learning, Fetoscopy, Rigid Registration and Video Mosaicking
Abstract:
Fetoscopic Laser Photocoagulation (FLP) is used to treat Twin-to-twin transfusion syndrome, however, this procedure is hindered because of difficulty in visualizing the intraoperative surgical environment due to limited surgical field-of-view, unusual placenta position, limited maneuverability of the fetoscope and poor visibility due to fluid turbidity and occlusions. Fetoscopic video mosaicking can create an expanded field-of-view (FOV) image of the fetoscopic intraoperative environment, which may support the surgeons in localizing the vascular anastomoses during the FLP procedure. However, existing classical video mosaicking methods tend to perform poorly in vivo fetoscopic videos. We propose the use of transformer-based detector-free local feature matching method as a dense feature matching technique for creating reliable mosaics with minimal drifting error. Using the publicly available fetoscopy placenta dataset, we experimentally show the robustness of the proposed method over the state-of-the-art vessel-based fetoscopic mosaicking method.
Detector-Free Dense Feature Matching for Fetoscopic Mosaicking