Reproducibility of liver R2* quantification for liver iron quantification from cardiac R2* acquisitions

To evaluate the reproducibility of liver R2* measurements between a 2D cardiac ECG-gated and a 3D breath-hold liver CSE-MRI acquisition for liver iron quantification.

A total of 54 1.5 T MRI exams from 51 subjects (18 women, 36 men, age 35.2 ± 21.8) were included. These included two sub-studies with 23 clinical MRI exams from 19 patients identified retrospectively, 24 participants with known or suspected iron overload, and 7 healthy volunteers acquired prospectively. The 2D cardiac and the 3D liver R2* maps were acquired in the same exam. Either acquisitions were reconstructed using a complex R2* algorithm that accounts for the presence of fat and residual phase errors due to eddy currents. Data were analyzed using colocalized ROIs in the liver.

Linear regression analysis demonstrated high Pearson's correlation and Lin's concordance coefficient for the overall study and both sub-studies. Bland-Altman analysis also showed good agreement, except for a slight increase of the mean R2* value above ~ 400 s^-1. The Kolmogorow-Smirnow test revealed a non-normal distribution for (R2* 3D-R2* 2D) values from 0 to 600 s^-1 in contrast to the 0-200 s^-1 and 0-400 s^-1 subpopulations. Linear regression analysis showed no relevant differences other than the intercept, likely due to only 7 measurements above 400 s^-1.

The results demonstrate that R2*-measurements in the liver are feasible using 2D cardiac R2* maps compared to 3D liver R2* maps as the reference. Liver R2* may be underestimated for R2* > 400 s^-1 using the 2D cardiac R2* mapping method.