Use of AI in Gastrointestinal Imaging

Medicine and Health Panel, 29 February 2024

Despite a rainy February evening we managed a good turnout including a healthy mixture of Medicine & Health (M&H) Panel members, wider WCIT and Livery members and their guests. Attendees were treated to an entertaining and thought-provoking presentation from Professor Gina Brown of Imperial College London covering the latest insights on the use of AI in gastrointestinal imaging to deliver better patient and healthcare outcomes.

Professor Brown is a leading expert from Imperial College with a distinguished career in cancer imaging (including 400+ papers published!). She holds a pivotal leadership role in national cancer imaging programmes and has also contributed enormously to wider international organisations.

Professor Brown explained some of the challenges in simply using AI on images in her field. Detection rates were not nearly as good compared to the standard practice. This involves two expert clinicians working independently followed by consensus meetings for any conflicting interpretations. Researchers found that by using a combination of AI and one expert initially, followed by the other expert providing validation, not only were the best diagnosis results obtained but some clinician time was also saved.

Another problem highlighted was if the scan data (including the additional patient and biopsy data )was not recorded in a consistent way. Gina has contributed significantly in this field by driving the standardisation of data collected via standard forms both in the UK and internationally. This is an important step if future advances are to be made using analysis of existing data.

Other complexities in the field were also identified. For example, existence of lymph nodes in scans was previously considered an indicator of poor survival rates but in many cases the opposite can be true. Existence of tumours connected to a vascular system is also an important indicator for staging and prognosis. Not only can these images be complex to analyse but the meaning and interpretation requires expertise currently too multi-layered for an AI system.

Early experiences with AI interpretation for breast cancer has shown some positive results in identification but has also led to too many false positives. This can result in more diagnoses and a greater workload on clinicians and resources.

Overall Prof Brown indicated several areas where future technology could improve the clinicians and patients lives. Examples included managing the patient experience and workflows. Auto population of reporting templates, improved diagnosis and research outcomes from better data collection and storage and better management of the security and privacy aspects were all areas needing improvement. Plenty of work and challenges going forward!

A lively Q&A session followed with good interaction from all the floor.  Freeman Rosey Woodhead provided an entertaining summary of the event in her vote of thanks and even managed to include a reference to Professor Brown’s passion for the Japanese art of Niwaki!

Everyone had a good chance to mix and chat afterwards and we all walked away with a far better understanding of the challenges and reality of trying to apply Technology and AI in this vital area.

A big thank you to Professor Brown for her time and energy, to the Hall team (especially Mellissa) and the Panel team (especially Rosey) for their enormous help in ensuring the success of this event.