Multi Texture Analysis of Colorectal Cancer Continuum Using Multispectral Imagery

• DOI: 10.1371/journal.pone.0149893
• Publication Date: 2016-02-22T19:15:19Z
• Authors: Ahmad Chaddad; Christian Desrosiers; Ahmed Bouridane; Matthew Toews; Lama Hassan; Camel Tanougast
• Keywords: Texture (cosmology)

Purpose This paper proposes to characterize the continuum of colorectal cancer (CRC) using multiple texture features extracted from multispectral optical microscopy images. Three types pathological tissues (PT) are considered: benign hyperplasia, intraepithelial neoplasia and carcinoma. Materials Methods In proposed approach, region interest containing PT is first images active contour segmentation. then encoded based on Laplacian-of-Gaussian (LoG) filter, discrete wavelets (DW) gray level co-occurrence matrices (GLCM). To assess significance textural differences between types, a statistical analysis Kruskal-Wallis test performed. The usefulness evaluated quantitatively in terms their ability predict various classifier models. Results Preliminary results show significant for all (p-value < 0.01). Individually, GLCM outperform LoG DW type prediction. However, higher performance can be achieved by combining features, resulting mean classification accuracy 98.92%, sensitivity 98.12%, specificity 99.67%. Conclusions These demonstrate efficiency effectiveness characterizing CRC discriminating