Joseph Mason Undergraduate Dissertation 2017/18

The Application of Machine Learning on MRI Data for Brain Disease Analysis

Supervised by H.Lu

Abstract

A software tool to improve the work-flow of experts analysing brain scans, by making use of unsupervised machine learning techniques. Human analysis of MRI data is time consuming and requires experts to look through many obtained scans with no order to their search. The project aims to produce a software tool to sort the relevance of brain scans by using a consensus clustering technique and the Connectome brain network data structure. The tool will offer a easy to use GUI front end to enable experts to make use of machine learning techniques without the need for technical skill. A novel relevance function will be suggested to interpret the most relevant scans in comparison to the currently observed scan. The GUI will update the list of relevant scans to help the expert in the task of classification.