Emine Ülkü Sarıtaş and Tolga Çukur and colleagues have published “Physics-Driven Autoregressive State Space Models for Medical Image Reconstruction” in IEEE Transactions on Medical Imaging (July 2026).
The paper introduces MambaRoll, an AI model that reconstructs MRI images from fast, incomplete scans. It combines the physics of how MRI data are acquired with a state space model that captures both fine detail and wider image context at low computational cost. Faster scans with good image quality can shorten exam times and make MRI more comfortable for patients.
Authors: Bilal Kabaş, Fuat Arslan, Valiyeh Ansarian Nezhad, Kader K. Oguz, Şaban Öztürk, Emine Ülkü Sarıtaş, Tolga Çukur