A machine learning pipeline for membrane segmentation of cryo-electron tomograms

Li Zhou, Chao Yang, Weiguo Gao, Talita Perciano, Karen M. Davies, Nicholas K. Sauter

Abstract: Highlights•Developed a novel machine learning pipeline using a convolutional neural network.•Segmentation pipeline uses reinforcement learning, region merging, and Gaussian processes.•New method targets membrane segmentation of cryo-electron tomograms.•Domain knowledge is used to generate training data.•The proposed fitting process in the pipeline also uses priors and constraints.•Demonstrated effectiveness and accuracy of the pipeline using two sets of real data.