Digital Health Journal Club | AI-Driven Automated Generation of Radiology Reports from Chest X-Rays

This month's Digital Health Journal Club session will be lead by Associate Professor Bevan Koopman, who will present on developing AI methods for automated generation of radiology reports from chest X-ray images.

Reporting on chest X-rays is a considerable overhead to radiologists; so much so that in some countries many X-rays are never reviewed by a radiologist. We aim to aid radiologists with AI tools that reduce the burden on reporting by generating a report directly from an x-ray image.

Key to our approach is alignment to real clinical workflow, where multiple images (often representing different views of the chest) are interpreted in conjunction by the model, as well as incorporating historical data about the patient, which all lead to more accurate report generation.

This session will present some technical details of image-to-text models in an accessible way as well as considering practical opportunities and barriers of how AI-driven automated reporting could be brought to the bedside. 

Recording available here.