Noninvasive classification of hepatic fibrosis based on texture parameters from double contrast‐enhanced magnetic resonance images

Abstract

PURPOSE: To demonstrate a proof of concept that quantitative texture feature analysis of double contrast-enhanced magnetic resonance imaging (MRI) can classify fibrosis noninvasively, using histology as a reference standard. MATERIALS AND METHODS: A Health Insurance Portability and Accountability Act (HIPAA)-compliant Institutional Review Board (IRB)-approved retrospective study of 68 patients with diffuse liver disease was performed at a tertiary liver center. All patients underwent double contrast-enhanced MRI, with histopathology-based staging of fibrosis obtained within 12 months of imaging. The MaZda software program was used to compute 279 texture parameters for each image. A statistical regularization technique, generalized linear model (GLM)-path, was used to develop a model based on texture features for dichotomous classification of fibrosis category (F ≤2 vs. F ≥3) of the 68 patients, with histology as the reference standard. The model's performance was assessed and cross-validated. There was no additional validation performed on an independent cohort. RESULTS: Cross-validated sensitivity, specificity, and total accuracy of the texture feature model in classifying fibrosis were 91.9%, 83.9%, and 88.2%, respectively. CONCLUSION: This study shows proof of concept that accurate, noninvasive classification of liver fibrosis is possible by applying quantitative texture analysis to double contrast-enhanced MRI. Further studies are needed in independent cohorts of subjects.