Multispectral biopsy image based colorectal tumor grader

Abstract

Automated tumor cell grading systems have an immense potential in improving the speed and accuracy of cancer diagnostic procedures. It can boost the confidence level of pathologists who perform the manual assessment of tumor cells. The application of image processing and machine learning techniques on the digitized biopsy slides enables the discrimination between various cell types. Deployment of multispectral imaging technique for biopsy slide digitization serves to provide spectral information along with the spatial information. Multispectral imaging allows to acquire several images of the sample in multiple wavelengths including the infrared ranges. This paper presents a multispectral image based colorectal tumor grading system. The algorithm validation is performed on our biopsy image database comprising 200 samples from 4 classes, viz. normal, hyperplastic polyp, tubular adenoma low grade as well as carcinoma cells. In addition to the visible bands, we have incorporated the spectral bands in near infrared ranges. Rotation invariant Local phase quantization (LPQ) feature extraction on our multispectral images have yielded a classification accuracy of 86.05% with an SVM classifier. Moreover, the experiments were carried out on another small multispectral image dataset which had 3 categories of cells.