Utility of Hepatic Perivascular Adipose Tissue as a CT Biomarker for Staging Liver Fibrosis

Repeated injury to the liver can cause liver fibrosis with end -stage cirrhosis being the 12th leading cause of death in the United States. Earlier stages of liver fibrosis can be reversed with treatment, but later stages are irreversible. A redistribution of liver segment volume and increase in periportal fat are previously known biomarkers of later fibrosis stages. We hypothesize that perivascular adipose tissue (PVAT) around the hepatic arteries and portal vein may also be predictive of liver fibrosis. In this pilot work, an automated pipeline was developed to segment the liver, spleen, and hepatic vessels on CT. Next, the PVAT around the hepatic arteries and portal vein was identified. Then, imaging-based biomarkers (e.g., volume, attenuation, fat fraction) of the liver, spleen, and PVAT were calculated. The biomarkers were used to train uni- and multi-variate logistic regression models to distinguish advanced fibrosis and cirrhosis. On a sample of 480 patients, the best multivariate model for cirrhosis achieved 96.3% AUC with biomarkers of hepatic PVAT, liver and spleen, and without blood serum scores. For advanced fibrosis, the multivariate model obtained 92.2 % AUC with features of PVAT, spleen and liver, and serum scores. To our knowledge, we are the first to show that PVAT-based biomarkers may be clinically useful for staging fibrosis.