Deep Learning Applied to Multi-Structure Segmentation in 2D Echocardiography: A Preliminary Investigation of the Required Database Size

Abstract

With the recent advances in machine learning, and their successful application to medical imaging, building medical databases big enough to learn to solve corresponding tasks for any given patient has become a priority. In this study, we set up a specific dataset of 500 patients to investigate the number of patients needed by two learning methods, the Structured Random Forest (SRF)and U-net, to obtain accurate segmentation results in 2D echocardiography. Our findings advocate that U-net is a good candidate to solve the still-open 2D echocardiography automatic segmentation problem.