Imaging is becoming a critical component of modern medicine; however, manual image analysis is time-consuming and resource-intensive.
The European Research Council (ERC) has awarded Bernhard Kainz a Consolidator Grant, recognizing his MIA-NORMAL project about automated medical image analysis as one of Europe's most prestigious research endeavours. The grant, worth two million euros over five years, aims to advance the development of machine intelligence-based tools that can accurately distinguish healthy human tissue from medical images. This will not only improve the efficiency and cost-effectiveness of medical diagnostics, but also provide essential support to overburdened medical professionals.
Imaging is becoming a critical component of modern medicine; however, manual image analysis is time-consuming and resource-intensive. With the integration of machine learning, Bernhard Kainz and his team aim to create computer programs that can pre-sort images into "probably healthy" or "possibly diseased" categories, saving medical staff valuable time to focus on more complex cases. The ultimate decision, however, remains with the medical staff.
Bernhard Kainz is driven by the belief that everyone deserves equal access to quality medical care, regardless of location or income. He and his team are developing methods to make high-quality medical image analysis accessible and scalable. Machine learning models will be trained to recognize healthy tissue structures rather than diagnose disease as it would take an immense amount of time and resources to train many different machine learning models on every possible disease.
The team aims to provide machine learning-based diagnostic tools in the future that can independently recognize what healthy anatomy should look like and continuously learn over time. The tools will detect normal physiological changes and unusual changes in individual patients, while also correlating patient information obtained by doctors with existing image material. In doing so, deviations requiring further medical attention can be identified and unnecessary examinations can be prevented.
Article text (excluding photos or graphics) © Imperial College London.
Photos and graphics subject to third party copyright used with permission or © Imperial College London.
Mr Ahmed Idle
Department of Computing