Fractal analysis differentiation of nuclear and vascular patterns in hepatocellular carcinomas and hepatic metastasis

Abstract

Hepatocellular carcinoma (HCC) currently represents the fifth most common cancer worldwide, while being the third leading cause of cancer death. Fractal analysis is a novel tool used in quantitative and qualitative image assessment. Vascular patterns and cellular nuclei particularities in tumoral pathology make ideal candidates for this technique. Our aim was to apply fractal analysis in quantifying nuclear chromatin patterns and vascular axels in order to identify differences between images of primary HCC, liver metastasis (LM) and surrounding normal liver tissue. Formalin-fixed, paraffin-embedded tissue sections from 40 cases of HCC and 40 LM of various origins were used. We performed Hematoxylin staining for nuclear chromatin as well as immunohistochemical staining for vascular patterns. High-resolution images were captured; nuclear and vascular morphologies were assessed on binarized skeleton masks using the fractal box counting method. Analysis was performed using the free, public domain Java-based image processing tool, ImageJ, which provided the fractal dimensions (FDs) for each studied element. Statistical analysis was performed using the ANOVA test with Bonferroni post-tests and t-tests for paired samples. Fractal analysis of vascular patterns clearly differentiated between tumoral tissue and normal surrounding tissue (p<0.01). Further analysis of nuclear FDs improved the specificity of these results, providing clear differentiation between pathological and normal tissue (p<0.01). When comparing primary HCC images with metastatic formations, we encountered statistically significant differences in nuclear chromatin assessment. However, blood vessels had a higher FD in primary tumors when compared with liver metastasis (p<0.05) and also allowed for a differentiation between primary liver tumors with and without neurodifferentiation. Fractal analysis represents a potent tool for discriminating between tumoral and non-tumoral tissue images. It provides accurate, quantifiable data, which can be easily correlated with the pathology at hand. Primary and metastatic liver tissue can be differentiated to some extent, however further studies, possibly including other variables (cellular matrix for instance) are needed in order to validate the method.