Quantitative versus morphological assessment of liver fibrosis: semi-quantitative scores are more robust than digital image fibrosis area estimation.

Abstract

BACKGROUND/AIM:

Digital image analysis (DIA) allows quantitative assessment of fibrosis on liver biopsy. Accurate determination of a threshold greyscale level representing fibrous tissue is critical. This method has not been fully evaluated in clinical practice.

METHODS:

Digital images of stained liver biopsy sections were captured by microscopy and converted to greyscale. A novel method of determining the threshold greyscale value at which to measure fibrosis area was developed (peak proportion area change (PPAC)). Reproducibility was tested in comparison with standard interactive thresholding and with semi-quantitative scoring using the Histological activity index (HAI) system by a histopathologist. Fibrosis areas for different sections from the same biopsy core were also compared by each method.

RESULTS:

Comparison between PPAC and interactive thresholding method demonstrated superior reproducibility of the PPAC method: r > 0.7, P < 0.001 compared with r = 0.19-0.64 (not all reaching significance). On a single section, reproducibility was similar for PPAC and the modified HAI system. When different sections from the same core were compared, the HAI system was more robust.

CONCLUSIONS:

The PPAC method is superior to standard interactive thresholding. However, variability in DIA scores between sections invalidates the technique for clinical use and semi-quantitative scoring systems remain the gold standard for fibrosis assessment.

• Authors: Mark Wright, Mark Thursz, Rebecca Pullen, Howard Thomas, Robert Goldin
• Journal: Liver international : official journal of the International Association for the Study of the Liver (Liver Int) Vol. 23 Issue 1 Pg. 28-34 (Feb 2003) ISSN: 1478-3223 [Print] United States
• PMID: 12640724 (Publication Type: Comparative Study, Evaluation Study, Journal Article)

Topics

• Fibrosis
• Hepatitis C, Chronic (pathology)
• Humans
• Image Processing, Computer-Assisted
• Liver (pathology)
• Observer Variation
• Reproducibility of Results