Click here to


Are you sure ?

Yes, do it No, cancel

The Linearity and Bias of Proton Density Fat Fraction (PDFF) Quantification at MRI Using a Micro-Emulsion Phantom

H Cho1*, C Lee1, B Ahn1, S Seok2, C Hong3, (1) Korea Research Institute of Standards and Science, Daejeon, Korea, (2) Hanyang University Medical Center, Seoul, Korea, (3) Daegu Catholic University, Daegu, Korea


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Fat quantification is important index of quantitative diagnosis, grading, and therapy monitoring of hepatic steatosis. However, the accuracy and reproducibility of fat fraction measurements varies according to the imaging system version, magnetic field strength, and reconstruction method. In this study, the consistency of fat fraction values was evaluated according to the imaging system version by using micro-emulsion phantom.

Methods: The surfactant-free micro-emulsion of various fat fractions, 22.9%, 39.4%, 55.4%, and 75.7%, were fabricated by using high-intensity focused ultrasonic emulsification technique. The true fat fraction of micro-emulsion was analyzed by Karl Fisher Titrator method. All emulsions were well-characterized and monitored for stability and accuracy. The six vials containing distilled water, grapeseed oil, and four micro-emulsions were prepared. The physical phantom consists of six vials surrounded by agarose gel. The emulsion phantom was scanned on Philips Achieva 3.0 T and Ingenia 3.0 T MRI (Philips Healthcare, Netherlands). The fat fraction was measured by the mDIXON-Quant which is the specific sequence for the proton density fat fraction (PDFF) measurement. The PDFF maps were generated across different MRI versions. The linearity and bias between MR imaging-PDFF and chemically analyzed fat fraction were evaluated.

Results: No significant artifacts were appeared on PDFF map. The PDFF values from two MRI systems were linear with the true fat fraction (R2 > 0.99). Regression slope (0.97) and intercept (3.48) of B system indicated slight overestimation. The mean Bland-Altman bias between two MRI system versions was -2.35 %. The limits of agreement were - 4.95 and +0.26.

Conclusion: The results show that the inconsistency of PDFF between two MRI systems. The system customized fat fraction correction based on the stable reference phantom will be essential to improve the consistency across sites, venders, field strengths and protocols.


MRI, Quantitative Imaging, Phantoms


Not Applicable / None Entered.

Contact Email