Fig. 1

Overview of the development of the deep-learning system, its workflow, and the test results. Subfigure (a) depicts the development process of the deep-learning system, which involves two steps: training a deep learning network to segment the single cardiac cycle regurgitation spectrum, and training another deep learning network to evaluate the severity of the regurgitation spectrum. Subfigure (b) outlines the workflow of the deep-learning system, indicating it as an end-to-end system. Subfigure (c) presents the results of the validation study, incorporating both internal dataset from our hospital and external dataset from seven other hospitals.