Artificial Intelligence-Guided Catheter Design for Patient-Specific Coronary Vein Procedure

Overview

CT scans are used in the medical field to identify a patient's anatomical features non -invasively. The coronary sinus ostium is a vital structure used in procedures for coronary venous system access but poses a challenge to physicians with its varied and small anatomy. Artificial Intelligence segmentation models and additional programs could utilize CT scans to create a 3D printable catheter sheath that can assist physicians in these procedures. A dataset was compiled of 4789 CT scans from the LIDC-IDRI with annotated cardiac structures from the Superior Vena Cava to the Coronary Sinus Ostium. A YOLOv11 Extra Large Segmentation Model was trained from scratch on this dataset for 778 epochs. This mo del achieved a mAP50-95 score of 0.67656 and F1 Score of 0.93. These statistics demonstrate the model’s high accuracy. A Python Application was curated by this study. It features the trained AI model identifying and segmenting the structures that the catheter will pass through. This data is analyzed for plausible points and a smoother algorithm was used to remove sharp ver tices. The program outputs a .stl file that could be used to 3D print. Statistics were collected and shown that the catheter had appropriate length and clearance from cardiac structures. This study creates a cutting-edge technology that can be utilized to assist physicians in coronary sinus ostium related procedures. Further training of the model, a practicality assessment to determine catheter materials and confirm use case, as well as cl inical trials are required before actual implementation in the field. West Virginia