Force Analysis Using Self-Expandable Valve Fluoroscopic Imaging: a way Through Artificial Intelligence.

Abstract

This study aimed to develop a force analysis model correlating fluoroscopic images of self-expandable valves with stress distribution. For this purpose, a nonmetallic measuring device designed to apply diverse forces at specific positions on a valve stent while simultaneously measuring force magnitude was manufactured, obtaining 465 sets of fluorescent films under different force conditions, resulting in 5580 images and their corresponding force tables. Using the XrayGLM, a mechanical analysis model based on valve fluorescence images was trained. The accuracy of the image force analysis using this model was approximately 70% (50-88.3%), with a relative accuracy of 93.3% (75-100%). This confirms that fluoroscopic images of transcatheter aortic valve replacement (TAVR) valve stents contain a wealth of mechanical information, and machine learning can be used to train models to recognize the relationship between stent images and force distribution, enhancing the understanding of TAVR complications.