Deep Learning for Inference of Hepatic Proton-Density Fat Fraction From T1-Weighted In-Phase and Opposed-Phase MRI: Retrospective Analysis of Population-Based Trial Data.

MedStar author(s):
Citation: AJR. American Journal of Roentgenology. 2023 Jul 19PMID: 37466189Institution: MedStar Health Research InstituteForm of publication: Journal ArticleMedline article type(s): Journal ArticleSubject headings: IN PROCESS -- NOT YET INDEXED | Year: 2023ISSN:
  • 0361-803X
Name of journal: AJR. American journal of roentgenologyAbstract: Background: The confounder-corrected chemical-shift-encoded MRI (CSE-MRI) sequence used to determine proton-density fat fraction (PDFF) for hepatic fat quantification lacks wide availability. Hepatic fat can alternatively be assessed using a 2-point Dixon method to calculate signal fat fraction (FF) from conventional T1-weighted (T1W) in-and-opposed-phase (IOP) images, although signal FF is prone to biases leading to inaccurate quantification. Objective: To compare hepatic fat quantification between PDFF inferred from conventional T1W IOP images using deep-learning convolutional neural networks (CNNs) and 2-point Dixon signal FF, using CSE-MRI PDFF as reference standard. Methods: This study entailed retrospective analysis of 292 participants (mean age, 53.7+/-12.0 years; 89 men, 103 women) enrolled at two sites from September 1, 2017 to December 18, 2019 in the Strong Heart Family Study (a prospective population-based study of American Indian communities). Participants underwent liver MRI (site A: 3 T; site B: 1.5 T) including T1W IOP MRI and CSE-MRI (used to reconstruct CSE-PDFF and CSE-R2* maps). Using CSEPDFF as reference, a CNN was trained in 218 (75%) randomly selected participants to infer voxel-by-voxel PDFF maps from T1W IOP images; testing was performed in the remaining 74 (25%) participants. Parametric values from the entire liver were automatically extracted. Per-participant median CNN-inferred PDFF and median 2-point Dixon signal FF were compared with reference median CSE-MRI PDFF using linear regression analysis, intraclass correlation (ICC), and Bland-Altman analysis. Code is publicly available: https://github.com/kang927/CNN-inference-of-PDFF-from-T1w-IOP-MR. Results: In the 74 test-set participants, reference CSE-PDFF ranged from 1% to 32% (mean, 11.3%+/-8.3%); reference CSE-R2* ranged from 31 to 457 s-1 (mean, 62.4+/-67.3 s-1). Agreement metrics with reference CSE-PDFF for CNN-inferred PDFF were: ICC=0.99, bias=-0.19%, 95% limits of agreement (LoA)=[-2.80%, 2.71%], and for 2-point Dixon signal FF were: ICC=0.93, bias=-1.11%, LoA=[-7.54%, 5.33%]. Conclusion: Agreement with reference CSE-PDFF was better for CNN-inferred PDFF using conventional T1W IOP images than for 2-point Dixon signal FF. Further investigation is needed in individuals with moderate-to-severe iron overload. Clinical Impact: CNN-inferred PDFF using widely available T1W IOP images may facilitate adoption of hepatic PDFF as a quantitative biomarker for liver fat assessment, expanding opportunities to screen for hepatic steatosis and non-alcoholic fatty liver disease.All authors: Ali T, Cole SA, Cunha GM, Hasenstab K, Henderson WC, Hsiao A, Middleton MS, Sirlin CB, Umans JG, Wang KFiscal year: FY2024Digital Object Identifier: Date added to catalog: 2023-08-15
No physical items for this record

Background: The confounder-corrected chemical-shift-encoded MRI (CSE-MRI) sequence used to determine proton-density fat fraction (PDFF) for hepatic fat quantification lacks wide availability. Hepatic fat can alternatively be assessed using a 2-point Dixon method to calculate signal fat fraction (FF) from conventional T1-weighted (T1W) in-and-opposed-phase (IOP) images, although signal FF is prone to biases leading to inaccurate quantification. Objective: To compare hepatic fat quantification between PDFF inferred from conventional T1W IOP images using deep-learning convolutional neural networks (CNNs) and 2-point Dixon signal FF, using CSE-MRI PDFF as reference standard. Methods: This study entailed retrospective analysis of 292 participants (mean age, 53.7+/-12.0 years; 89 men, 103 women) enrolled at two sites from September 1, 2017 to December 18, 2019 in the Strong Heart Family Study (a prospective population-based study of American Indian communities). Participants underwent liver MRI (site A: 3 T; site B: 1.5 T) including T1W IOP MRI and CSE-MRI (used to reconstruct CSE-PDFF and CSE-R2* maps). Using CSEPDFF as reference, a CNN was trained in 218 (75%) randomly selected participants to infer voxel-by-voxel PDFF maps from T1W IOP images; testing was performed in the remaining 74 (25%) participants. Parametric values from the entire liver were automatically extracted. Per-participant median CNN-inferred PDFF and median 2-point Dixon signal FF were compared with reference median CSE-MRI PDFF using linear regression analysis, intraclass correlation (ICC), and Bland-Altman analysis. Code is publicly available: https://github.com/kang927/CNN-inference-of-PDFF-from-T1w-IOP-MR. Results: In the 74 test-set participants, reference CSE-PDFF ranged from 1% to 32% (mean, 11.3%+/-8.3%); reference CSE-R2* ranged from 31 to 457 s-1 (mean, 62.4+/-67.3 s-1). Agreement metrics with reference CSE-PDFF for CNN-inferred PDFF were: ICC=0.99, bias=-0.19%, 95% limits of agreement (LoA)=[-2.80%, 2.71%], and for 2-point Dixon signal FF were: ICC=0.93, bias=-1.11%, LoA=[-7.54%, 5.33%]. Conclusion: Agreement with reference CSE-PDFF was better for CNN-inferred PDFF using conventional T1W IOP images than for 2-point Dixon signal FF. Further investigation is needed in individuals with moderate-to-severe iron overload. Clinical Impact: CNN-inferred PDFF using widely available T1W IOP images may facilitate adoption of hepatic PDFF as a quantitative biomarker for liver fat assessment, expanding opportunities to screen for hepatic steatosis and non-alcoholic fatty liver disease.

English

Powered by Koha