Deep learning for risk stratification of patients with liver cancers

Lead Supervisor
Professor Julia Schnabel
Professor of Computational Imaging
School of Biomedical Engineering and Imaging Science
julia.schnabel@kcl.ac.uk

Co-supervisor
Dr Cheng Fang
Consultant Radiologist, King’s College Hospital

Project Details

Liver disease is the third leading cause of premature death in the UK. The liver cancer incidence rate is projected to rise by 38% in the UK between 2014 and 2035. There are around 6000 cases of primary liver cancer diagnosed each year. In addition, 600,000 people have some form of liver disease in England and Wales and are at increased risk of developing liver cancer.  Medical imaging is paramount in the management of chronic liver disease and liver cancer in terms of screening, diagnosis, deciding treatment strategy and monitoring treatment response. Currently there is no reliable non-invasive methods/model of predicting biological nature of the tumour. Decision on treatment strategy has been largely based on experts opinion, tumour size and tumour location.  A more accurate assessment for tumour risk stratification is required. 

The aim of the project is to use deep learning technique to predict the treatment response and likelihood of treatment relapse, leading to a more accurate risk stratification prior to treatment. The project will involve identify a retrospective dataset from patients with imaging and clinical follow-ups at key clinical milestones and prepare the dataset for deep learning. Building/optimising convolutional convolution networks (CNN) using the dataset as the initial training and validation data sets.

Datasets

Retrospective imaging and clinical data from KCH. We will be identifying a cohort of patients with initial diagnostic cross-sectional imaging as well as post-treatment follow-up imaging. The usage of the data will be subject for R&D approval. 

Keywords

Deep learning, CNN, liver cancer, prognosis, biomarkers