Magn Reson Med. 2025 Oct 28. doi: 10.1002/mrm.70162. Online ahead of print.
ABSTRACT
PURPOSE: To create virtual liver models simulating the concurrent presence of hepatic iron overload and steatosis based on human liver histology and to assess the accuracy and precision of R2* and fat fraction (FF) estimation at 0.55T compared to 1.5T.
METHODS: Virtual liver models simulating the coexistence of iron overload and steatosis were created, encompassing clinical liver iron concentration (LIC) (0-20 mg/g) and FF (0%-30%) ranges, using morphological characteristics derived from human liver histology. MRI signals were synthesized via Monte Carlo simulations under varying SNR conditions (SNR original (SNRorig) and SNR enhanced (SNRenh)) at 0.55T (SNRorig,0.55T = 20; SNRenh,0.55T = 50) and 1.5T (SNRorig,1.5T = 50). Simulation results were validated using iron-fat phantoms (FF: 0%-40%; iron concentration: 0-8 μg/mL) scanned across varying SNRs using different signal averages (SNRorig,0.55T, SNRorig,1.5T = 1; SNRenh,0.55T = 4). Accuracy and precision of 0.55T for estimating R2* and FF compared to 1.5T were assessed using boxplots, heatmaps, and Kruskal-Wallis statistical tests.
RESULTS: SNRorig,0.55T demonstrated reduced or comparable FF estimation bias relative to SNRorig,1.5T, with significantly lower bias for SNRenh,0.55T, especially for severe iron overload and FF ≤ 20%. LIC estimation for SNRorig,0.55T and SNRenh,0.55T demonstrated higher accuracy compared to SNRorig,1.5T with significant differences (p < 0.028) for all combinations of LIC and FF except for severe iron overload with FF > 10% (p > 0.183). Further, SNRenh,0.55T showed improved precision for FF and LIC estimation compared to SNRorig,1.5T. Phantom experiments showed reduced bias for both FF and LIC estimation for SNRenh,0.55T compared to SNRorig,1.5T.
CONCLUSION: 0.55T shows promising results for concurrent R2* and FF estimation, enhancing accuracy and precision compared to 1.5T, especially with enhanced SNR.
PMID:41152137 | DOI:10.1002/mrm.70162
