Systematic quantification of histological patterns shows accuracy in reflecting cirrhotic remodeling.

ABSTRACT

BACKGROUND AND AIM: There still lacks a tool for precisely evaluating cirrhotic remodeling. Histologic distortion characterized in cirrhosis (i.e. cirrhotic patterns) has a validated pathophysiological meaning and potential relevance to clinical complications. We aimed to establish a new tool to quantify the cirrhotic patterns and test it for reflecting the cirrhotic remodeling. METHODS: We designed a computerized algorithm, named qCP, dedicated for the analysis of liver images acquired by second harmonic microscopy. We evaluated its measurement by using a cohort of 95 biopsies (Ishak staging F4/5/6 = 33/35/27) of chronic hepatitis B and a carbon tetrachloride-intoxicated rat model for simulating the bidirectional cirrhotic change. RESULTS: QCP can characterize 14 histological cirrhosis parameters involving the nodules, septa, sinusoid, and vessels. For chronic hepatitis B biopsies, the mean overall intra-observer and inter-observer agreement was 0.94 ± 0.08 and 0.93 ± 0.09, respectively. The robustness in resisting sample adequacy-related scoring error was demonstrated. The proportionate areas of total (collagen proportionate area), septal (septal collagen proportionate area [SPA]), sinusoidal, and vessel collagen, nodule area, and nodule density (ND) were associated with Ishak staging (P < 0.01 for all). But only ND and SPA were independently associated (P ≤ 0.001 for both). A histological cirrhosis parameters-composed qCP-index demonstrated an excellent accuracy in quantitatively diagnosing evolving cirrhosis (areas under receiver operating characteristic curves 0.95-0.92; sensitivity 0.93-0.82; specificity 0.94-0.85). In the rat model, changes in collagen proportionate area, SPA, and ND had strong correlations with both cirrhosis progression and regression and faithfully characterized the histological evolution. CONCLUSIONS: QCP preliminarily demonstrates potential for quantitating cirrhotic remodeling with high resolution and accuracy. Further validation with in-study cohorts and multiple-etiologies is warranted.