AI and the World's Hard Problems · Medicine & drug discovery · arXiv · 2026-10-20
GliomaINK: A Tumor and Anatomy Conditioned Flow Matching Framework for 3D MRI Sequence Synthesis in Glioma Patients
GliomaINK is a conditional flow matching framework for synthesizing missing 3D brain MRI sequences in glioma patients, with an anatomy-guided loss to preserve structure. Trained on 1,083 patients from BraTS-AG, it achieved the highest PSNR across all six MRI translation directions in both test cohorts. For T2W-to-T1C synthesis, PSNR reached 25.64 dB, compared with 18.90 dB for SPADE and 13.78 dB for Pix2Pix.