ABSTRACT
Colorectal Adenocarcinoma is the most primarily universal category of Colon Cancer. It basically originates in the intestinal glandular structures. Thus, in clinical practices, to forecast and map its course of cure, the intestinal glands’ morphology along with architectural structure and glandular development information is used by the pathologists. A lot of digital automated techniques are proposed on regular basis for removing the need for manual grading and providing better accuracy. However, achieving accurate cancer grading still remains a big challenge for modern pathology. In order to curb this challenge, an automated supervised technique using deep learning keeping original image size is proposed in this paper for doing five-grade cancer classification via 31 layers deep CNN. The proposed model results classification accuracy of 96.97% for two-class grading and 93.24% for five-class cancer grading.
