Deep preferential evolution guided by pairwise comparisons enables motion correction in photoacoustic tomography
Teaching AI to fix blurry medical images by comparing which looks better
Researchers created a new method to fix motion blur in photoacoustic tomography—a medical imaging technique that maps tissue using sound waves and light. Instead of telling an algorithm what a perfect image should look like, they trained it to simply compare two images side-by-side and pick the better one, then used that judgment to guide a search algorithm toward clearer pictures. The approach worked on real human images even though it was mainly trained on simulations, and it corrected motion blur from natural, unpredictable body movement.
Photoacoustic tomography could enable cheaper, more portable medical imaging, but motion during scanning ruins the pictures. Current methods for fixing blur rely on mathematical rules that don't match how doctors actually judge image quality. This pairwise-comparison approach produces sharper, more useful images from motion-corrupted scans, and because it learns from comparisons rather than absolute scores, it transfers better to real patient data—meaning fewer perfectly labeled training images are needed to deploy it in clinics.