Samuel Remedios, a PhD student in the Department of Computer Science, received a Best Poster Award at the 2025 Information Processing in Medical Imaging conference for his work, “Cycle-Consistent Zero-Shot Through-Plane Super-Resolution for Anisotropic Head MRI.” The conference took place May 25 through 30 on Kos Island in Greece.
Remedios’s work addresses a common clinical reality: Many MRIs are acquired as 2D slices and stacked into 3D volumes, producing crisp images in the slice plane, but blurred views from other angles. Radiologists can work around this, but AI systems expect truly 3D data.
Remedios’s method turns low-detail medical scans into clearer, 3D images that computers can analyze, with two built-in guardrails: First, it constrains the enhanced image so that the result is a faithful representation of the original data collected in the scan; second, a trained AI keeps the final image consistent with real brain anatomy.
Presented as a poster, the project stood out for its rigor and open discussion, the judges said.
Next, Remedios plans to expand his method beyond one type of brain scan to work with different kinds of medical scans and other body parts by training his system on new data.