AI supported colorectal surgery triage
Supervisors: Professor Evropi Theodoratou (cancer epidemiologist), Dr Athina Spiliopoulou (data scientist), Professor Malcolm Dunlop (consultant colorectal surgeon)
Colorectal cancer (CRC) is the third most common cancer across the world, with 1.8 million cases and ~860,000 deaths each year. Survival of CRC patients depends on early detection. Therefore, identifying factors that reliably determine the existence of colorectal neoplasia and related pathologies from available data sources is vital for early detection and accurate referral systems. This PhD project will evaluate existing and develop new machine learning methods, with an emphasis in deep neural networks architectures to study and determine the predictive value of clinical, laboratory and pathologic factors in people with suspected CRC. The resulting AI methodology will mine relevant patient data from electronic health systems, such as the NHS TRAK, and will be used to triage referrals of patients with GI symptoms to surgery and gastrointestinal clinics.
A pilot system has been developed for NHS Lothian Gastroenterology by Ian Arnott’s (Consultant Gastroenterologist) in collaboration with Deloitte to triage referrals to gastroenterology. The proposal is to adapt and extend this model to triage referrals to colorectal surgery, in collaboration with the NHSL Colorectal Surgeons, led by Prof Malcolm Dunlop and Dr Farhat Din and NHSL gastroenterologists, led by Mr Ian Arnott. The proposal will use natural language processing methodology (BERT models and recurrent neural networks) to triage based on text in the referral letter. The academic collaboration is in place with access to the necessary data.