Novel application of the gray-level co-occurrence matrix analysis in the parvalbumin stained hippocampal gyrus dentatus in distinct rat models of Parkinson's disease

Abstract

To reveal the best choice of algorithm for parvalbumin-immunostained images of the hippocampal gyrus dentatus in two distinct rat models of Parkinson's disease (PD), particularly in terms of extracting the crucial information from the image, we tested whether the impact of experimentally induced dopaminergic (hemiparkinsonism) vs. cholinergic (PD cholinopathy) innervation impairment on the parvalbumin stained GABA interneurons could be detected using two separate algorithms, the fractal box-count and the gray-level co-occurrence matrix analysis (GLCM) algorithms. For the texture and fractal analysis of the hippocampal gyrus dentatus images, we used.tif images from three experimental groups of adult male Wistar rats: control rats, rats with Parkinson disease (PD) cholinergic neuropathology (with a PPT lesion), and hemiparkinsonian rats (with a SNpc lesion). For the suprapyramidal layer of the gyrus dentatus ASM and Entropy differentiated the images of the SNpc lesion versus the images of the control and the PPT lesion subjects, with significantly higher ASM and lower Entropy, indicating the homogenization of the images and their lower gray-level complexity. The infrapyramidal images of the SNpc group were differentiated versus the images from the control and PPT groups in terms of all the GLCM parameters: they showed lower mean Entropy and Contrast and higher ASM, Correlation and IDM. These results strongly suggest a rise in the uniformity, homogeneity and orderliness in the gray-levels of images from the SNpc group. Our results indicate that GLCM analysis is a more sensitive tool than fractal analysis for the detection of increased dendritic arborization in histological images.