Standardized Approach for ROI-Based Measurements of Proton Density Fat Fraction and R2 in the Liver

• Published in: American journal of roentgenology (1976) Vol. 209; no. 3; pp. 592 - 603
• Main Authors: Campo, Camilo A., Hernando, Diego, Schubert, Tilman, Bookwalter, Candice A., Pay, Andrew J. Van, Reeder, Scott B.
• Format: Journal Article
• Language: English
• Published: United States 01.09.2017
• Subjects: Adolescent; Adult; Aged; Aged, 80 and over; Case-Control Studies; Child; Female; Humans; Image Interpretation, Computer-Assisted; Iron Overload - diagnostic imaging; Magnetic Resonance Imaging - methods; Male; Middle Aged; Non-alcoholic Fatty Liver Disease - diagnostic imaging; Protons; Reproducibility of Results; Tomography, X-Ray Computed; R2; proton density fat fraction; liver disease; MRI; ROI
• ISSN: 0361-803X; 1546-3141; 1546-3141
• DOI: 10.2214/AJR.17.17812

The purpose of this study was to evaluate the reproducibility (interreviewer agreement) and repeatability (intrareviewer agreement) of ROI sampling strategies to measure chemical shift-encoded (CSE) MRI-based liver proton density fat fraction (PDFF) and R2* (1 / T2*). A secondary purpose was to standardize ROI-based liver PDFF and R2* measurements by providing a compromise between measurement reproducibility and repeatability and time burden for image analysts. CSE data from two cohorts were retrospectively analyzed. Cohort A included 53 patients referred for abdominal MRI and healthy subjects recruited for a comparison study of CT and MRI. Cohort B included 37 patients with suspected liver iron overload. Three reviewers measured liver PDFF and R2* using previously reported ROI sampling strategies. Inter- and intrareviewer agreement of liver PDFF and R2* were evaluated using Bland-Altman analysis. Averaging largest-fit ROIs over the nine Couinaud segments resulted in the narrowest limits of agreement (LOA) for liver PDFF and R2* measurements in both cohorts. For PDFF, interreviewer agreement had mean LOA of ± 0.8% for cohort A and ± 1.7% for cohort B. Intrareviewer agreement was ± 0.5% for cohort A and ± 0.9% for cohort B. For R2* interre-viewer agreement had mean LOA of ± 3.0 s for cohort A and ± 17.9 s for cohort B. Intrare-viewer agreement was ± 2.6 s for cohort A and ± 14.6 s for cohort B. This approach was the most time-burdensome, requiring a mean ± SD of 149.7 ± 8.6 s per dataset. For improved reproducibility and repeatability of liver PDFF and R2* measurements, clinicians and researchers should sample as much area of the liver as possible using multiple large ROIs.