Deep learning tissue segmentation of beating embryonic hearts from 4-D OCT images

Optical coherence tomography (OCT) has been applied to investigate heart development because of its capability to image both structure and function of tiny beating embryonic hearts. Labeling heart structures is necessary for quantifying mechanical functions such as cardiac motion, wall strain, blood flow and shear stress, of looping hearts. Since manual segmentation is time-consuming and labor- intensive, this study aimed to use deep learning to automatically extract dynamic shapes including the myocardium, the endocardial cushions, and the lumen of beating embryonic hearts from 4-D OCT images. This will benefit research on heart development, especially studies requiring large cohorts of embryos.