Meeting Banner
Abstract #0102

Multi-Peak Spectral Modeling of Fat Is Necessary for Both Accurate Liver Fat & Iron Quantification: A Biopsy-MRI Correlation Study

Jens-Peter Khn1, 2, Diego Hernando1, Norbert Hosten2, Scott B. Reeder1, 3

1Department of Diagnostic Radiology, Wisconsin Institutes for Medical Research USA, Madison, WI, United States; 2Department of Diagnostic Radiology, University Greifswald, Greifswald, MV, Germany; 3Medical Physics, Biomedical Engineering and Medicine, Wisconsin Institutes for Medical Research USA, Madison, WI, United States

Chemical shift-encoded MRI methods provide measures of liver fat, and also R2*, a biomarker for hepatic iron content. Previous works have shown that correcting for the spectral complexity of fat (multi-peak modeling) is necessary for accurate fat quantification, however it is unknown whether it is necessary for accurate iron quantification. In this work, we investigated the impact of multi-peak fat modeling on R2* measurements to assess hepatic iron content, using biopsy as reference standard. In the presence of fat, single-peak fat modeling results in clinically significant errors in R2* quantification. Multi-peak fat modeling removes these errors and is necessary for accurate hepatic iron quantification.

Abstract PDF Presentation Video

Keywords

abdominal abnormal accelerator according accounting accuracy accurate acknowledge acknowledgments addition additionally affected aged agreed apparent bias biomedical biopsy black blue board breath brunt capable cells certified chemical circles classified clear clinical clinically closely coefficient complexity confirmed confounding containing content core corrected correction correlation defined density dependence determined diagnosis diagnostic disagreement distinguish droplets engineering errors established estimation evaluated evaluating foundation fraction fractions generated good grade graded grades gradient grading granules guided hepatic histological hold hours impact inaccurate incorrect increasing indicated indicating institutes intercept intracellular invasive investigate iron lead least literature liver location mapping maps measure measured measuring medical mild model modeling models moderate necessary noise operator overload part pathologists patients peak percentage peter physics presence proton quantification radiologist radiology reader reconstructed reconstruction recorded reeder removes reported requires respectively samples scale segment sensitivity separation several slope specificity spectral staining steps strong support system systematically threshold underestimated underestimation underwent unknown water whether women years