Stem Cell Classification With Convolutional Neural Networks

Abstract: In this bachelor thesis project, the problem of imageclassification with convolutional neural networks is considered.In several fields of biology, automatized cell detection is a helpfultool for facilitating the process of cellular analysis. This reportanswers the question whether a computer program can tell if animage contains muscle stem cells or not. Analogously to the neuronsof the human brain, the creation of such a program involvestraining thousands of mathematically modeled artificial neuronsto maximize the likelihood of producing correct classifications.This report covers how such a network is implemented and showshow its performance dependens on the network’s dimensions. It isrevealed that a neural network indeed can replace and speed upthe manual process of classifying images. With an image datasetof cells, the best performing networks manage to classify imageswith an accuracy of up to 90%.