Classification of Colorectal Cancer Based on Histological Image Using a Combination of Color Histogram, Haralick and k-NN

• Published in: 2022 International Conference on Information Technology Research and Innovation (ICITRI) pp. 179 - 183
• Main Authors: Wildah, Siti Khotimatul, Agustiani, Sarifah, Latif, Abdul, Pebrianto, Rangga, Hasan, Fuad Nur, Indriyani, Fintri
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10-11-2022
• Subjects: Color Histogram; Colorectal Cancer; Feature extraction; Haralick; Histograms; History; Image color analysis; Intestines; k-NN; Machine Learning; Technological innovation; Training data
• DOI: 10.1109/ICITRI56423.2022.9970214

Colorectal cancer is the progression of malignancy of cancer that occurs in the colon or rectum, which is a small part of the large intestine before the anus. Colorectal cancer is the third most common type of malignant cancer in the world, and it is estimated that by 2030 it will increase by 60%, from 2.2 million new cases with a prevalence of 1.1 million deaths from this cancer. Colorectal cancer can occur due to several factors such as a history of colon polyps, a history of chronic diseases that occur in the intestines and lifestyle and often men are more at risk of developing the disease than women. In the early stages of cancer often cannot be known and do not cause symptoms and will only be realized after developing quite severe, therefore an early examination is needed in order to determine the condition of the cancer and as an anticipation or prevention so as not to worsen the situation. Examination of colorectal cancer can be done with a biopsy technique, where tissue or histopathological images can be observed. can be done automatically by utilizing computer vision technology using machine learning methods. This study aims to propose a method that can classify colorectal cancer 5000 histology images. The method proposed in this study uses the Random Forest machine learning method with a combination of Color Histogram and Haralick feature extraction. The results obtained in this study provide a good accuracy value of 93,53%.