Researchers in Canada Combine Physics, AI to Clarify Medical Images for Eye Disease Diagnosis

A research team at the University of Waterloo, Ontario, has unveiled a new artificial intelligence (AI) approach that aims to improve the clarity and detail of eye images used for disease diagnosis. By teaching AI software the physics of how light travels through eye tissue, the scientists have created a model capable of restoring the fine details essential for detecting corneal diseases.

The newly developed Physics-Informed Diffusion Model (PIDM) is trained not just on data, but also on the underlying physics of how light interacts with tissue. By understanding how defocus and speckle noise are formed, the model can reverse image quality loss and progressively refine OCT scans, verifying each step against real-world physics, according to a University of Waterloo news release.

“Typical diffusion AI models can sometimes misinterpret or ‘hallucinate’ details when reconstructing images,” said Dr. Alexander Wong, professor of systems design engineering and Canada Research Chair in Medical Imaging Systems. “By merging the power of AI with the knowledge of physics, our model methodically reduces such errors and produces more trustworthy results.”

In tests on both plant tissue and human corneal scans, the PIDM outperformed existing reconstruction techniques, revealing crisp cellular outlines and intricate internal structures.

“This technology comes at a time when OCT imaging of the eye is becoming more common and will be crucial to its widespread adoption by eyecare practitioners worldwide. It could enable earlier detection of external eye diseases and help catch problems that might otherwise be missed," said Dr. Lyndon Jones, Principal Scientist at the Centre for Ocular Research and Education.

The research team, which also includes Nima Abbasi, a PhD candidate in systems design engineering, plans to incorporate additional physics principles to extend the technology beyond the cornea, with future applications aimed at retinal imaging and other ocular tissues.

The full study, "A Physics-Informed Diffusion Model for Super Resolved Reconstruction of Optical Coherence Tomography Data," is published in IEEE Transactions on Biomedical Engineering.