This AI spots dangerous blood cells doctors often miss

Researchers developed CytoDiffusion, a generative AI system for analyzing blood cells, with the aim to enhance diagnoses of diseases such as leukemia.
CytoDiffusion utilizes advanced AI technology to examine not just typical patterns but subtle variations in blood cell appearances, allowing for more sensitive detection of abnormalities.
Unlike other AI tools focused on predefined categories, CytoDiffusion models the full range of blood cell features, enhancing its ability to detect rare or unusual cells.
The system was trained on an extensive dataset of over half a million blood smear images from Addenbrooke’s Hospital, making it robust against variations in different medical settings.
CytoDiffusion demonstrated superior ability in identifying leukemia cells compared to current systems and human specialists, showing greater confidence and accuracy in uncertain situations.
Researchers released the world’s largest public collection of blood smear images to facilitate further development and validation of AI tools for global research communities.
CytoDiffusion is designed to assist clinicians in managing large datasets and flagging concerning cases, augmenting, rather than replacing, human expertise in medical diagnostics.