Single cell classification : An approach to understand the diversity of protein localization

Abstract

A deep learning approach for the classification of protein localizations in microscopic biological cell images is presented here. A pretrained segmentation network is utilized to extract single cells from cell population images. A convolutional neural network (CNN) is then trained on the extracted cells. The CNN was able to achieve an F1 score of 0.42 on the training dataset and 0.39 on the validation dataset. The results show that the network can classify some protein localizations with higher accuracy than others.