What type of study is this?

September 17, 2025

Clinical Validation of Deep-learning Approach to Estimate Proton-Density Fat Fraction from In-phase and Out-of-phase T1-weighted MRI

Key Points

AI-IOP PDFF closely matched CSE-PDFF, suggesting high accuracy for liver fat quantification.
The deep-learning method outperformed conventional Dixon SFF in liver fat estimation accuracy.
External validation confirmed the reliability of AI-IOP across diverse settings, indicating robustness.
Findings may enhance diagnosis of metabolic-associated steatohepatitis where advanced methods are inaccessible.

Abstract

Motivation: Accurate liver fat quantification is vital for diagnosing Metabolic-associated Steatohepatitis (MASH). Conventional IOP MRI faces limitations, and advanced CSE methods are not widely available. Goal(s): Evaluate AI-IOP PDFF externally, compare its accuracy with 2-point Dixon SFF, and identify factors influencing PDFF estimation for model improvement. Approach: Analyzed 526 liver MRIs with CSE-MRI and T1w-IOP sequences. Generated CSE-PDFF, Dixon SFF, and AI-IOP PDFF maps using automated segmentation and statistical analyses. Results: AI-IOP PDFF closely matched CSE-PDFF and outperformed Dixon SFF in accuracy. External validation confirmed its reliability across settings. Impact: AI-IOP enables reliable liver fat quantification with common T1w-IOP sequences, enhancing MASH diagnosis where advanced CSE methods are unavailable.

Clinical Validation of Deep-learning Approach to Estimate Proton-Density Fat Fraction from In-phase and Out-of-phase T1-weighted MRI

Key Points

Abstract

Cite This Study