SUMI: Distilling Photon-Counting CT into Routine Chest CT via Clinically Validated Degradation Modeling

Summary

Researchers Junqi Liu, Xinze Zhou, and Wenxuan Li present SUMI, a deep learning approach (arXiv:2504.06848, published April 9, 2025, cs.CV) that synthesizes photon-counting CT (PCCT) quality images from routine energy-integrating chest CT (EICT). PCCT offers higher spatial resolution and lower noise than conventional CT, but its limited clinical availability restricts large-scale research. SUMI learns to reverse realistic acquisition artifacts in low-quality EICT by modeling the degradation from PCCT to EICT in a clinically validated way. On external data, SUMI outperforms state-of-the-art enhancement methods by 15% in SSIM and 20% in PSNR, improves radiologist-rated clinical utility, and boosts downstream lesion detection performance. The work offers a practical path to bring PCCT-like image quality to hospitals using standard CT scanners.