Introduction

Quantitative T1, T2, T2* and fat-fraction (FF) maps are promising imaging biomarkers for the assessment and follow-up of diffuse liver diseases^1,2. Liver magnetic resonance fingerprinting (MRF)³ enables simultaneous co-registered quantification of T1, T2, T2*, and FF in a single breath-hold scan, suitable for non-invasive liver tissue characterization⁴. To progress along the pathway to imaging biomarker development, such techniques should be validated against standard techniques in patients and their measurement tied to pathological changes⁵.

Here we (i) evaluated the agreement between multi-parametric liver MRF-derived metrics and reference standard imaging in patients with diffuse liver disease, (ii) evaluated the scan-rescan repeatability of the liver MRF measurement, and (iii) validated MRF-derived measurements for histological grading of liver biopsies.

Methods

Study population
This prospective study was approved by the local institutional review board. Written informed consent was obtained from each subject prior to scanning. A total of 46 patients (19 men and 27 women; age, 59.8±15.0 [28–85] years) with diffuse liver disease undergoing MRI were enrolled in this study.

MR imaging and reconstruction
Liver MRF acquisition is based on a 9-echo, golden-angle radial gradient-echo (GRE) acquisition with bipolar readouts and varying inversion recovery and T2 preparation pulses, allowing simultaneous liver T1, T2, T2*, and FF mapping in a single ~20-s breath-hold scan^4,6. Each patient was imaged with a 9-echo liver MRF sequence and individual conventional mapping acquired in three separate breath-holds (T1 modified Look-Locker inversion recovery [MOLLI], T2 gradient spin echo [GraSE], and multiecho GRE, for T1, T2, T2*/FF estimation, respectively). Rescan of the 9-echo liver MRF was also performed at the end of the same session. All measurements were performed on a 1.5-T scanner (Ingenia; Philips Healthcare, Best, The Netherlands) with a 32 channel body coil. Offline reconstruction was performed in MATLAB (MATLAB2019a; MathWorks, Natick, MA) using dictionary-based sparse and locally low rank regularized reconstruction⁷.

Liver histology
Of the 46 patients, 10 patients (4 men and 6 women; age, 46.1±9.4 [29–58] years) underwent liver biopsy. Liver biopsies were percutaneously performed by a gastroenterologist through the intercostal space in the midaxillary line. Formalin-fixed biopsy specimens were stained with hematoxylin and eosin (for overall assessment), Azan trichrome stain (to assess fibrosis), and Berlin blue stain (to assess iron deposition). Biopsy specimens were evaluated by a hepatic pathologist (23 years of experience) blinded to both the MRF and conventional maps. Biopsies were scored using the NASH Clinical Research Network NAFLD activity score and fibrosis score⁸.

Analysis
For each parameter map, a total of ten circular regions of interest (ROIs) were manually drawn in the liver (four different areas avoiding large vessels), erector spinae muscle (one on each side), subcutaneous fat (one on each side), and the spleen (two different areas) by two board-certified radiologists (7 and 20 years of experience) blinded to the pathology. Median value was recorded for each ROI and compared to its corresponding reference measurements. Scan-rescan repeatability of each MRF parameter map was assessed using the coefficient of variation (CV), defined as the standard deviation (SD) of the quantitative maps derived from the scan and rescan over the mean of both maps. Welch's t-test with Bonferroni correction was used to compare MRF-derived metrics among the histological grades. The degree of association between continuous variables was calculated using the Pearson’s correlation coefficient.

Results

Comparison with reference standard imaging and repeatability of MRF metrics
Figure 1 shows representative MRF maps and conventional T1, T2, T2*, and FF maps of a patient with diffuse liver disease. Agreement with reference standard measurements in patients with diffuse liver disease is shown in Figure 2A (r=0.93, r=0.93, r=0.69, and r=0.99 for T1, T2, T2*, and FF, respectively). A slight T1 overestimation compared to MOLLI and underestimation of T2 were consistent with previous MRF findings⁶. The scan-rescan repeatability of each parameter acquired with MRF is shown in Figures 2B and 2C. The overall CVs for repeatability in the liver were 4%, 9%, and 10% for T1, T2, and T2*, respectively. Note that this repeatability includes errors due to poor breath-holding and patient movements.

Histopathological correlation
Representative maps and histology images of patients with severe steatosis and fibrosis are shown in Figures 3 and 4, respectively. MRF-derived metrics showed good agreement (r=0.85, r=0.95, and r=0.62 for fibrosis, steatosis, and inflammation, respectively) between histological evaluation of liver biopsies from patients with diffuse liver disease (Figure 5). After correcting for multiple comparisons, MRF showed significantly higher T1 values in patients with histologically proven fibrosis (P<0.01 for all combinations), higher FF in steatosis (P<0.01 for all combinations), and higher T1 values in inflammation (P<0.01 except for Grade 0 vs. 1, P=0.78).

Conclusion

Multi-parametric liver MRF provided repeatable T1, T2, T2*, and FF maps with high agreement with measurements of separate conventional quantitative mapping in patients with diffuse liver disease. Furthermore, measurements of liver MRF showed good agreement between the pathological grades of the liver biopsies. Future work will aim to increase the number of subjects with liver biopsy including siderosis as well as to assess longitudinal changes to monitor treatment response and disease progression.