The MPhys project I completed had 2 parts, a radiomics section and a machine learning section:
-
Radiomics is a large feature-based extraction technique. This part was initially to set up a workflow which included converting the NiFtI MRI images into a suitable format using Lua and MATLAB.
-
Deep learning was used (with a U-Net like convolutional neural network) in order to perform segmentation on the MRI images. This was done to identify the cancerous tumour regions using hand-drawn delineations as the labels.