Authors
Shin Hye Hwang, Sejin Jeong, Dong-Kyu Kim, Seungsoo Lee, Soo Yoon Chung, Seok-Jae Heo, Gyeonghyun Cho, Kyungchul Song, Byung Won Yoo, Eun Byoul Lee, Hyun Joo Shin
Published in
Ultrasonography (Seoul, Korea). Volume 45. Issue 6. Pages 618-628. Epub Oct 14, 2026.
Abstract
This study evaluated the diagnostic performance of the ultrasound-guided attenuation parameter (UGAP) for pediatric hepatic steatosis and developed a multivariable model to estimate magnetic resonance imaging proton density fat fraction (MRI-PDFF) from ultrasound and clinical variables.
This single-center retrospective study included pediatric patients who underwent UGAP measurement and MRI-PDFF assessment within 3 months of each other. UGAP performance in detecting moderate-to-severe (≥S2) and severe (S3) steatosis was evaluated using the area under the receiver operating characteristic curve (AUC). Multivariable models were developed using linear regression and a random forest algorithm with five-fold cross-validation.
Among 45 patients (33 male and 12 female; median age, 12 years), median UGAP and MRI-PDFF values were 0.83 dB/cm/MHz and 22%, respectively. UGAP correlated moderately with MRI-PDFF (ρ=0.596, P<0.001). The AUC was 0.74 for both ≥S2 and S3 steatosis (95% confidence intervals [CIs], 0.58 to 0.91 and 0.60 to 0.89, respectively), and two discordant cases were identified. A model incorporating UGAP, age, height, and liver size outperformed UGAP alone in detecting ≥S2 steatosis (optimism-corrected AUC, 0.882 vs. 0.744; P=0.01). In the machine learning analysis, the model showed stable performance during cross-validation, with an AUC of 0.789 (95% CI, 0.649 to 0.928) for ≥S2 steatosis and an R² of 0.663 for continuous MRI-PDFF prediction.
UGAP correlated moderately with MRI-PDFF and showed fair performance in grading pediatric hepatic steatosis. Combining UGAP with clinical parameters significantly improved MRI-PDFF estimation, providing a practical screening tool for hepatic steatosis when MRI is unavailable. External validation is needed.
PMID:
42958863
Bibliographic data and abstract were imported from PubMed on 30 Oct 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 15
- Comments 0