The development of a multi-stage learning scheme using new tissue descriptors for automatic grading of prostatic carcinoma

This paper introduces a new system for the automated classification of prostatic carcinomas from biopsy images. The important components of the proposed system are (1) the new features for tissue description based on hyper-complex wavelet analysis, quaternion color ratios, and modified local binary patterns; and (2) a new framework for multi-stage learning that integrates both multi-class and binary classifiers. The system performance is estimated by employing Hold-out cross-validation in a dataset of 71 prostate cancer biopsy images with different Gleason grades. Simulation results show that the presented technique is able to correctly classify images in 98.89% of the test cases. Furthermore, the system is robust in terms of sensitivity (0.9833) and specificity (0.9917). We have demonstrated the efficacy of our system in distinguishing between Gleason grades 3, 4 and 5.