Viability Analysis of Drug-Treated Tumor Spheroids Using Machine Learning

Abstract

3D spheroids that are able to mimic the microenvironment of tumors effectively have emerged as significant structures in cancer biology and drug development. This study aims to help cancer researchers monitor the changes in human liver cancer spheroids in response to drug treatment by offering a software tool for evaluating cell viability within 3D spheroids. A dataset of spheroid images are collected, processed, and classified using alternative machine learning models constructed with Random Forest, Logistic Regression, Support Vector Machine and Extreme Gradient Boosting methods. The classification performances of the models are evaluated in terms of the prediction accuracy, precision, recall, and F1-score values. Based on the test experiments conducted, Extreme Gradient Boosting model achieved the highest ratios for all of the performance metrics. Furthermore, a standalone desktop application is implemented to perform analyses of the images with the help of its user-friendly interface.