Premium
Accuracy and the effect of possible subject‐based confounders of magnitude‐based MRI for estimating hepatic proton density fat fraction in adults, using MR spectroscopy as reference
Author(s) -
Heba Elhamy R.,
Desai Ajinkya,
Zand Kevin A.,
Hamilton Gavin,
Wolfson Tanya,
Schlein Alexandra N.,
Gamst Anthony,
Loomba Rohit,
Sirlin Claude B.,
Middleton Michael S.
Publication year - 2016
Publication title -
journal of magnetic resonance imaging
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.563
H-Index - 160
eISSN - 1522-2586
pISSN - 1053-1807
DOI - 10.1002/jmri.25006
Subject(s) - nuclear medicine , confounding , magnetic resonance imaging , linear regression , multivariate statistics , medicine , body mass index , flip angle , voxel , in vivo magnetic resonance spectroscopy , reproducibility , multivariate analysis , mathematics , nuclear magnetic resonance , statistics , radiology , physics
Purpose To determine the accuracy and the effect of possible subject‐based confounders of magnitude‐based magnetic resonance imaging (MRI) for estimating hepatic proton density fat fraction (PDFF) for different numbers of echoes in adults with known or suspected nonalcoholic fatty liver disease, using MR spectroscopy (MRS) as a reference. Materials and Methods In this retrospective analysis of 506 adults, hepatic PDFF was estimated by unenhanced 3.0T MRI, using right‐lobe MRS as reference. Regions of interest placed on source images and on six‐echo parametric PDFF maps were colocalized to MRS voxel location. Accuracy using different numbers of echoes was assessed by regression and Bland–Altman analysis; slope, intercept, average bias, and R 2 were calculated. The effect of age, sex, and body mass index (BMI) on hepatic PDFF accuracy was investigated using multivariate linear regression analyses. Results MRI closely agreed with MRS for all tested methods. For three‐ to six‐echo methods, slope, regression intercept, average bias, and R 2 were 1.01–0.99, 0.11–0.62%, 0.24–0.56%, and 0.981–0.982, respectively. Slope was closest to unity for the five‐echo method. The two‐echo method was least accurate, underestimating PDFF by an average of 2.93%, compared to an average of 0.23–0.69% for the other methods. Statistically significant but clinically nonmeaningful effects on PDFF error were found for subject BMI ( P range: 0.0016 to 0.0783), male sex ( P range: 0.015 to 0.037), and no statistically significant effect was found for subject age ( P range: 0.18‐0.24). Conclusion Hepatic magnitude‐based MRI PDFF estimates using three, four, five, and six echoes, and six‐echo parametric maps are accurate compared to reference MRS values, and that accuracy is not meaningfully confounded by age, sex, or BMI. J. Magn. Reson. Imaging 2016;43:398–406.