Open Access

ARTICLE

Two Stage Classification with CNN for Colorectal Cancer Detection

Pallabi Sharma^1,*, Kangkana Bora², Kunio Kasugai³, Bunil Kumar Balabantaray¹

1 Department of Computer Science and Engineering, National Institute of Technology Meghalaya, Shillong, 793003, India
2 Computer Science and Information Technology, Cotton University, Guwahati, 781001, India
3 Department of Gastroenterology, Aichi Medical University, Nagakute, 480-1195, Japan

* Corresponding Author: Pallabi Sharma. Email:

Oncologie 2020, 22(3), 129-145. https://doi.org/10.32604/oncologie.2020.013870

Abstract

In this paper, we address a current problem in medical image processing, the detection of colorectal cancer from colonoscopy videos. According to worldwide cancer statistics, colorectal cancer is one of the most common cancers. The process of screening and the removal of pre-cancerous cells from the large intestine is a crucial task to date. The traditional manual process is dependent on the expertise of the medical practitioner. In this paper, a two-stage classification is proposed to detect colorectal cancer. In the first stage, frames of colonoscopy video are extracted and are rated as significant if it contains a polyp, and these results are then aggregated in a second stage to come to an overall decision concerning the final classification of that frame to be neoplastic and non-neoplastic. In doing so, a comparative study is being made by considering the applicability of deep learning to perform this two-stage classification. The CNN models namely VGG16, VGG19, Inception V3, Xception, GoogLeNet, ResNet50, ResNet100, DenseNet, NASNetMobile, MobilenetV2, InceptionResNetV2 and fine-tuned version of each model is evaluated. It is observed that the VGG19 model is the best deep learning method for colonoscopy image diagnosis.

---

Keywords

---

Citations