This repository contains the master's thesis codes.
The project aims to predict the treatment response of Malignant Pleural Mesothelioma (MPM) patients with machine learning (ML) methods by using 3D CT scans.
In the first part, an unsupervised ML method was implemented on the features extracted from CTs using radiomics. Pyradiomics was used as software to extract features from CT scans. The result was compared with labels created by looking at the volume change on the follow-up CT scans.
In the second part, supervised ML methods, Convolutional Neural Network(CNN) based deep neural network models were employed. Tensorflow was used to create CNN classification, while the other models were created using Pytorch.