Digital Twins for Liver Cancer Using Medically Informed Machine Learning

This PhD research project develops digital twin models for the human liver for use in liver cancer treatment planning and optimisation. The project creates an image-based computational model of the liver with realistic anatomical variability, structure, and aspects of functionality. Digital twins have many use cases including training clinicians, testing computational algorithms, planning patient-specific treatments, and enabling virtual in-silico trials for evaluating novel treatments. The project focuses on hepatocellular carcinoma (HCC), the most common type of primary liver cancer in adults and currently the most common cause of death in people with cirrhosis. The research will develop a liver function map as part of a digital twin model using biomarkers from pre-operative MRI to estimate tissue-level inflammation, fibrosis, fat content, cirrhosis, and hepatocyte uptake extraction rate. This combines magnetic resonance imaging with medically informed machine learning (MIML) techniques. The project seeks to exploit multi-modal imaging (CT, MRI) and novel data-driven machine learning methods to develop and validate the digital twin model. The liver function maps will be paired with anatomical shape models of the liver and computational algorithms for generating the vasculature of the liver and the tumour, creating the first ever digital twin model of the human liver. MIML techniques will be used to reduce model overfitting and ensure consistency with existing medical knowledge about tumour pathophysiology, vascular function, and tissue response to radiotherapy.