Related Experiment Video
Updated: Apr 19, 2026
Fat-Water Phantoms for Magnetic Resonance Imaging Validation: A Flexible and Scalable Protocol
Published on: September 7, 2018
Reproducibility of ROI methods and reliability of conventional versus chemical shift-encoded MRI for hepatic fat and
Hilton Leao Filho1, Virginia Goncalves2, Nathalia Mendonca2
1Radiology Department, Hospital do Coração, São Paulo, Brazil. hleao@hcor.com.br.
Purpose:
To compare the inter-reader reproducibility of two ROI placement strategies (circular versus freehand geographic) for hepatic PDFF and R2* quantification, and determine R2* thresholds above which conventional techniques show increasing disagreement relative to confounder-corrected chemical shift-encoded (CSE) methods, used as the reference technique.
Materials And Methods:
In this retrospective study, consecutive adults underwent abdominal MRI at 1.5 T with CSE and conventional sequences. Four radiologists (two seniors, two fellows) independently measured PDFF and R2* using both strategies. Inter-reader agreement was assessed using ICC; conventional-CSE agreement was evaluated with MAE stratified by R2* thresholds.
Results:
Excellent interreader agreement was observed for all strategies: PDFF ICCs were 0.974 (five-ROI mean) and 0.975 (geographic); R2* ICCs were 0.983 for both. Correlation between strategies was near-perfect (PDFF: r = 0.995; R2*: r = 0.991). No significant differences were found between experience levels. A sliding threshold analysis identified conventional R2* = 45 s⁻¹ as a proposed exploratory cutoff (bootstrap 95% CI: 40-70 s⁻¹): below this value, conventional PDFF maintained low error (MAE = 3.2% points [pp]; 17.9% exceeding 5 pp), whereas above 80 s⁻¹, MAE increased to 11.2 pp with 93.8% exceeding 5 pp. Conventional R2* systematically overestimated iron with increasing PDFF (MAE = 12.8 s⁻¹ at PDFF ≥ 20%). In this study, reproducibility refers to inter-reader consistency (ICC), whereas agreement denotes error relative to the CSE reference (MAE).
Conclusion:
Both ROI strategies yield excellent, comparable reproducibility regardless of reader experience. A proposed exploratory three-zone framework using conventional R2* (GREEN < 45 s⁻¹, YELLOW 45-80, RED ≥ 80) provides preliminary guidance regarding when conventional PDFF shows lower error relative to the CSE reference and when CSE methods should be strongly considered, pending external validation.
