This refers to the LogitGaze-Med project, developed at Innopolis University. The system can predict how a doctor examines an X-ray image. The project improved the accuracy of pathology detection by up to 5%.
The model can analyze not only images but also medical context, taking into account anatomical structures, opacities, heart contours, and other key features.
What does the technology enable?
- Predicting a radiologist's gaze trajectory;
- Identifying problematic areas in the image;
- Explaining medical AI decisions;
- Improving diagnostic accuracy for pneumonia and heart failure.
The developers believe that the technology will help make medical AI systems more understandable and increase doctors' trust in such solutions.