3rd Year Dissertation

This was my third year computer science project during which we have to develop an artefact and discuss everything from its rationale to the processes used for its conception. My chosen project was based around the theme of using machine learning techniques to accurately help diagnose and define brain tumours. There is a lot of detail contained within the document approximately 20,000 words so you may wish to just skim it.

The role machine learning could play in helping to diagnose Brain tumours and grade them appropriately according to the World Health Organisation Standard is extraordinary. The successful development and subsequent application of a learning algorithm that could assist in diagnosing Brain Tumours, minimising the need for surgical biopsies required to perform histopathological analysis on the suspect tumours. This studies aims to explore the application of Support Vector Machines and Artificial Neural Networks to a collection of 1st and 2nd order statistical features collected from regions of interest processed from FLAIR MRI images (25). With the feature selection performed using a Decision Tree learning approach (CART) to create the smallest possible hypothesis. The accuracy was measured by the use of a validation set using leave-out-p cross validation. Two SVM were built and a one ANN with moderate results of 76% classification accuracy.

Abstract from the document

TitleMRI brain tumour grading using machine learning on statistical features

UniversityUniversity of Lincoln

DescriptionThe role machine learning could play in helping to diagnose Brain tumours and grade them appropriately according to the World Health Organisation Standard is extraordinary. The successful development and subsequent application of a learning algorithm that could assist in diagnosing Brain Tumours, minimising the need for surgical biopsies required to perform histopathological analysis on the suspect tumours. This studies aims to explore the application of Support Vector Machines and Artificial Neural Networks to a collection of 1st and 2nd order statistical features collected from regions of interest processed from FLAIR MRI images (25).