Whole-liver histogram and texture analysis on T1 maps improves the risk stratification of advanced fibrosis in NAFLD

Objectives: To assess whole-liver texture analysis on T1 maps for risk stratification of advanced fibrosis in patients with suspected nonalcoholic fatty liver disease (NAFLD).

Methods: This retrospective study included 53 patients. Histogram and texture parameters (volume, mean, SD, median, 5th percentile, 95th percentile, skewness, kurtosis, diff-entropy, diff-variance, contrast, and entropy) of T1 maps were calculated based on the semi-automatically segmented whole-liver volume. A two-step approach combining the Nonalcoholic Fatty Liver Disease Fibrosis Score (NFS) and Fibrosis-4 Index (FIB-4) with the liver stiffness measurement (LSM) for the risk stratification was used. Univariate analysis was performed to identify significant parameters. Logistic regression models were then run on the significant features. Diagnostic performance was evaluated with receiver operating characteristic (ROC) analysis.

Results: In total, 33 (62%) subjects had a low risk and 20 (38%) subjects had an intermediate-to-high risk of advanced fibrosis. The following significantly different parameters with the best performance were diff-entropy, entropy, and diff-variance, with AUROC 0.837 (95% CI 0.73-0.95), 0.821 (95% CI 0.71-0.94), and 0.807 (95% CI 0.69-0.93). The optimal combination of median, 5th percentile, and diff-entropy as a multivariate model improved the diagnostic performance to diagnose an intermediate-to-high risk of advanced fibrosis with AUROC 0.902(95% CI 0.79-0.97).

Conclusions: Parameters obtained by histogram and texture analysis of T1 maps may be a noninvasive analytical approach for stratifying the risk of advanced fibrosis in NAFLD.

Key points: • Variable flip angle (VFA) T1 mapping can be used to acquire 3D T1 maps within a clinically acceptable duration. • Whole-liver histogram and texture parameters on T1 maps in patients with NAFLD can distinguish those with an intermediate-to-high risk of advanced fibrosis. • The multivariate model of combination of texture parameters improved the diagnostic performance for a high risk of advanced fibrosis and clinical parameters offer no added value to the multivariate model.