3D Ultrafast ICE Image-based In-vivo Ablation Catheter Tip Detection

Intracardiac echocardiography (ICE) is the leading imaging modality for various cardiac interventions, providing real-time images of the anatomical structures and interventional tools such as ablation catheters. Image-based localization of the ablation catheter can reduce procedure time and exposure to fluoroscopy. We propose an ablation catheter tip detection framework that combines a 3D nnU-Net-based segmentation and post-processing. Our method allows us to robustly detect the ablation catheter tip. The proposed method achieved a segmentation test dice score of 0.80 and 0.78, mean tip location error of 2.1 mm and 2.1 mm, and mean tip angular error of 2.4° and 2.3°, when trained with Delay-and-Sum (DAS) and Iterative Maximum a Posteriori (IMAP) beamformed images, respectively.