A new generative AI system developed by researchers from the University of Cambridge, University College London and Queen Mary University of London can analyse blood cell images with high accuracy, identifying rare abnormalities that may signal diseases such as leukaemia.

• Technical Approach: The AI model was trained on a dataset of peripheral blood smear images, enabling it to recognise the full range of normal and pathological cell appearances.
• Performance: In evaluations, the system detected subtle signs of disease with sensitivity equal to or exceeding current leading models and greater consistency than human specialists.
• Uncertainty Awareness Unlike conventional models, the system can assess and communicate its own uncertainty when encountering ambiguous or unfamiliar cell patterns.

This summary is based on a ScienceDaily article reporting research from the University of Cambridge.