Bringing eye images into focus with AI

At the University of Waterloo, researchers have developed a new tool combining physics and artificial intelligence to improve the clarity of eye imaging used to diagnose disease. The new AI model, a physics-informed diffusion model (PIDM), is trained on both image data and the physics of how light interacts with tissue. This allows it to reverse image degradation such as defocus and speckle noise in scans of the cornea. In tests using plant tissues and human corneal scans captured by optical coherence tomography (OCT), the model produced much clearer images, revealing detailed cellular structures that traditional methods could not. By providing sharper, more reliable eye images, this technology could enable earlier and more accurate diagnosis of external eye diseases. Led by Professor Alexander Wong, the research builds on the legacy of the late Professor Kostadinka Bizheva, who dedicated her career to advancing high-resolution eye imaging.