Fully Automated High-Throughput Quantitative MRI of the Liver

Confirm the accuracy (ie: bias) of the proposed CSE-MRI method in patients with liver steatosis and in patients with liver iron overload

For each liver segment, and for whole-liver PDFF and R2\* measurements, we will determine bias of CSE 2D using CSE 3D BH as the reference from Bland-Altman analysis. The range of R2\* values leading to reliable measurements of PDFF and R2\* will be determined using a two-segment piecewise linear model with the change point estimated from the data.

Image quality of free breathing 2D chemical shift encoded (CSE) method

Image quality for each acquisition will be evaluated by 3 radiologists, using a Likert scale between 0 (worst/non-diagnostic) and 3 (best) for criteria including: motion artifacts, spatial resolution, SNR (signal- to-noise ratio), and overall image quality.

Measure the repeatability of the proposed CSE-MRI method in patients with liver steatosis and in patients with liver iron overload

To compare repeatability for various acquisition/reconstruction methods, we will calculate the squared difference between each pair of repeated PDFF / R2\* measurements, and will model this value including the acquisition/reconstruction method as a covariate using Generalized Estimating Equations (GEE), with subject as a random factor to account for within-subject correlation. This analysis will be performed for all liver segments and whole-liver measurements. Further, the repeatability coefficient will also be assessed separately for each method. Bland-Altman plots will also be used to visualize the repeatability profile over the PDFF / R2\* range. The automated vs manual PDFF / R2\* measurements will be evaluated for each acquisition.

Confirm the reproducibility of the proposed CSE-MRI method across field strengths

The reproducibility of the proposed method between 1.5T and 3T will be evaluated in two ways. First, at each field strength, PDFF and R2\* will each be compared between FB FAM CSE and BH CSE 3D (as the reference), using Bland Altman analysis to assess the bias. Second, PDFF will be compared between 1.5T and 3T using Bland Altman analysis for each method, based on the expected field strength independence of PDFF. R2\* will also be compared between 1.5T and 3T for each method, after converting to liver iron concentration (LIC, mg/g) using the 1.5T and 3T calibrations determined by our group, to compensate for the effect of field strength on R2\*. In all comparisons, the reproducibility coefficient, and interclass correlation coefficient (ICC) with 95% confidence intervals will be calculated to quantify the degree of reproducibility.

Technical success rate of the integrated single-button push, fully automated CSE-MRI protocol in < 5 minutes of MR room time

The technical success rate of the integrated single-button push, fully automated CSE-MRI exam will be determined through completion rate of the exam in less than 5 minutes, radiologist evaluation of image quality using a Likert scale, and accuracy of the automated analysis compared to manual evaluation.