Chest / ThoracicAI / InformaticsResearch

RadiolGAN Synthetic CT from UTE MRI Improves Lung Image Quality and Diagnostic Confidence

Radiology AI literature (PubMed)today

RadiolGAN synthesized CT-like images from 3D UTE MRI, achieving higher PSNR (65.3 vs 64.7) and diagnostic confidence (3.98 vs 3.59) than CycleGAN, though GGO depiction similar (prospective multicenter, 359 subjects).

  • Prospective multicenter study in 359 subjects (4 centers; external test n=54).
  • RadiolGAN-CT outperformed CycleGAN-CT in PSNR (65.32 vs 64.68) and diagnostic confidence (3.98 vs 3.59), but SNR/CNR/noise did not differ significantly.
  • Limitation: Feasibility study with no formal diagnostic accuracy comparison against real CT; external test set relatively small.
Read the source

RadPigeon summaries are original and for information only. They are not clinical advice.