Deep Learning For Transesophageal Echocardiography View Classification

This study presents a deep learning model that can accurately classify eight commonly used intraoperative and intraprocedural TEE views across a wide range of clinical and echocardiographic characteristics. The model was tested on a combination of randomly selected internal and external test videos and demonstrated high accuracy in both standard black-and-white 2D B-Mode TEE videos and videos incorporating color flow Doppler information. This is the first application of a machine learning strategy to TEE video image data acquired during the course of standard clinical care for open cardiac surgeries and transcatheter procedures.