Purpose or Case Report: Compressed SENSE (CS) facilitates significantly shorter MRI acquisition times. There is a paucity of data comparing quantitative MRI with regularly under-sampling schemes like SENSE to pseudorandom under-sampling schemes like CS. The purpose was to compare measurements of liver proton density fat fraction (PDFF), T2*, and stiffness between SENSE and CS techniques with identical MR acquisition parameters.
Methods & Materials: Clinical liver MRI examinations performed between March 2019 and September 2019 that included quantitative MRI with both SENSE and CS were retrospectively identified. The CS acceleration factors were selected to allow a 50% breath-hold duration compared to our routine clinical SENSE acquisitions. Patient demographics were recorded, along with liver PDFF (%), T2* (ms), and stiffness (kPa). Weighted (by region of interest area) mean was computed for each parameter at four anatomic levels for both SENSE and CS acquisitions. Spearman’s rank-order correlation (r) was used to evaluate the association between conventional and CS measurements, and Bland-Altman plots were used to determine means bias and 95% limits of agreement (LOA).
Results: Thirty-three examinations met inclusion criteria. Mean patient age was 14.1 ± 4.6 years; eleven exams (33%) were in female patients. Liver PDFF showed very strong positive correlation (r=0.99) between sequences, with a mean bias of 0.26% (95% LOA: -0.91, 1.43%). Liver T2* showed weak positive correlation (r=0.37), with a mean bias of -3.2 ms (95% LOA: -12.44, 6.11 ms) that decreased with longer T2*, and it included two extreme outliers due to motion artifacts. Liver stiffness showed very strong positive correlation (r=0.97) with a mean bias of 0.1 kPa (95% LOA: -0.36, 0.56 kPa) that increased with increasing liver stiffness.
Conclusions: In our population, there was very strong correlation between SENSE and CS-based MRI measurements of liver PDFF and stiffness, with minimal bias. However, there was relatively poor correlation between T2* measurements, with a moderate bias between SENSE and CS under-sampling schemes.