AI-Assisted Medical Imaging Analysis Improves Diagnostic Accuracy

Strategic Insights -

Cardiff University researchers have developed an AI system that examines medical images like a trained radiologist, with the help of input and expertise from radiologists. The study combined radiologists' eye movements with other AI systems to improve diagnostic performance by up to 1.5%. This breakthrough could support decision-making in diagnoses and grow medical AI adoption to meet NHS challenges.

Key Takeaways:

  • The researchers created the largest and most reliable visual saliency dataset for chest X-rays to date, based on over 100,000 eye movements from 13 radiologists examining fewer than 200 chest X-rays.
  • The dataset was used to train a new AI model, CXRSalNet, to help it predict the areas in an X-ray that are most likely to be important for diagnosis.
  • The AI system improved diagnostic performance by up to 1.5% and aligned machine behavior more closely with expert human judgment.
  • The study aimed to bring together two aspects of radiology: the computer's ability to identify pathologies and the radiologist's knowledge of where to look on imaging studies.
  • The researchers plan to further develop their approach to explore how the technology can be adapted for medical education and training, and clinical decision support tools.

Statistics:

  • The dataset consisted of over 100,000 eye movements from 13 radiologists examining fewer than 200 chest X-rays.
  • The AI system improved diagnostic performance by up to 1.5%.
  • The UK has a 29% shortfall in consultant radiologists, according to the 2024 census by the Royal College of Radiologists.
  • Demand for imaging is rising significantly.
  • Wales has a 32% shortfall in consultant radiologists.

Sources:

  • Cardiff University
  • IEEE Transactions on Neural Networks and Learning Systems
  • Royal College of Radiologists
  • Cardiff University School of Computer Science and Informatics
  • University Hospital of Wales (UHW)
  • Hantao Liu, Professor of Human-Centric Artificial Intelligence, Cardiff University
  • Richard White, Consultant Radiologist at UHW and clinical lead on the study