Can Machine Learning Based Models Reliably Differentiate Between Aggressive and Non-Aggressive Tumours when Predicting Prostate Cancer Diagnosis using Novel Biomarkers found in Patient Biofluids?
Project Abstract
Prostate cancer annually effects around 52,000 people in the United Kingdom and is often diagnosed through invasive procedures. This project aims to prove that a machine learning based model will be more capable of differentiating between aggressive and non-aggressive tumours in prostate cancer diagnosis through the analysis of novel biomarkers in patient biofluids – reducing the need for invasive procedures. Rarely have studies been done that fully combine urological and computer science studies. This project will create a Random Forest algorithm to attempt to replicate previous positive findings in cancer research, and compare this against similar machine learning techniques such as decision trees and K-nearest neighbour algorithms to determine which technique is most efficient in answering the research question. Hopefully, the results will demonstrate how effective and reliable machine learning methods can be in clinical settings, and how we can move away from unreliable and invasive diagnostic criteria.
Keywords: Machine Learning, Cancer Research, Computational Biology
Conference Details
Session: Presentation Stream 8 at Presentation Slot 10
Location: GH037 at Tuesday 7th 13:30 – 17:00
Markers: Galileo Sator (GTA), Ulrich Berger
Course: MSc Computer Science, Masters PG
Future Plans: I’m continuing studies