Brian Thomas Tischler MD, Presenter: Nothing to Disclose

Naznin Daginawala MD, Abstract Co-Author: Nothing to Disclose

Karen Buch MD, Abstract Co-Author: Nothing to Disclose

Hei Shun Yu MD, Abstract Co-Author: Nothing to Disclose

Baojun Li PhD, Abstract Co-Author: Nothing to Disclose

Jorge A. Soto MD, Abstract Co-Author: Nothing to Disclose

Cindy Christiansen PhD, Abstract Co-Author: Nothing to Disclose

Stephan W. Anderson MD, Abstract Co-Author: Nothing to Disclose

The gold standard for diagnosing hepatic fibrosis is percutaneous biopsy; an invasive procedure with limitations and complications including sampling error, morbidity and mortality. Developing noninvasive approaches to diagnose fibrosis using imaging is therefore clinically important. Non-contrast CT (NCCT) is an imaging modality with several advantages when compared to some other imaging options as it has no contraindications and thus can be performed on nearly any patient regardless of their renal function, allergies, or internal ferromagnetic materials. The purpose of this study was to evaluate the ability of a texture analysis program to grade hepatic fibrosis on NCCT.

Following IRB approval, 59 patients with a random liver biopsy within 6 months of having a NCCT were included. Hepatic segmentation of 5 slices through the porta hepatis on each patient’s NCCT was performed, and an in-house developed MATLAB texture analysis program was used to extract 42 texture features. Ishak Fibrosis Scale (scores 0-6) was used to determine the biopsy specimens’ histopathologic fibrosis scores. A classification and regression tree (CART) analysis was performed to find texture features that most correlated with the hepatic fibrosis scores. Patients were separated into 2 groups: low level fibrosis 0-2 versus higher levels of fibrosis 3-6, and low-moderate levels 0-3 versus high levels of fibrosis 4-6.

Included patients’ fibrosis scores ranged from 0-6. CART analysis found short run emphasis (SRE), long run high gray-level emphasis (LRHGE), mean CT number, and 9 neighborhood standard deviation (Std9) to be the main texture features used to distinguish patients with low fibrosis scores 0-2 from higher fibrosis scores 3-6 with a sensitivity of 100%, specificity of 97% and PPV of 96%. CART analysis found Std5, LRHGE and Law’s feature 5 (L5) to be the main texture features used to distinguish low-moderate fibrosis scores 0-3 from high fibrosis scores 4-6 with a sensitivity of 88%, specificity of 98%, and PPV of 93%.  

This study shows that texture analysis of NCCT images can accurately distinguish low levels from higher levels of hepatic fibrosis.

Texture analysis of NCCT images is a potential alternative to liver biopsy for evaluating hepatic fibrosis because it is noninvasive, has no contraindications, and can accurately distinguish low levels from higher levels of hepatic fibrosis.

Tischler, B, Daginawala, N, Buch, K, Yu, H, Li, B, Soto, J, Christiansen, C, Anderson, S, Texture Analysis of Non-Contrast CT Images to Assess Hepatic Fibrosis.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14008877.html