SRGAN Reconstructs Ultrafast Synthetic MRI Quantitative Maps
Researchers at Beijing Friendship Hospital and Beihang University report in JMIR Med Inform (2026) that a superresolution generative adversarial network (SRGAN) can reconstruct ultrafast synthetic MRI into whole-brain T1, T2, and proton-density maps closely matching routine scans. In a prospective study of 151 healthy adults and seven patients, SRGAN outputs showed strong correlation (T1 R²=0.98; T2 R²=0.97; PD R²=0.99) and halved acquisition time while maintaining diagnostic image quality.
Key Points
- 1Shows SRGAN reconstructs T1/T2/PD maps with strong correlation (T1 R²=0.98, T2=0.97, PD=0.99)
- 2Reduces whole-brain acquisition time by approximately 50% while suppressing noise and artifacts in fast scans
- 3Enables clinically usable quantitative maps with small biases (T1 0.93%, T2 −0.85%, PD 0.31%), aiding faster diagnosis
Scoring Rationale
Strong experimental validation and clinical relevance, limited by single-center cohort and moderate systematic underestimation of T2 values.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
