Colorectal Cancer Histology Image Classification Using Stacked Ensembles

Abstract

—In this research, we propose the image processing and classification using image extraction and data mining with ensemble learning techniques. We apply image extraction to determine the appropriate subset of the initial features to be used later by the ensemble learning for inducing an accurate model to predict cancer from the colorectal cancer histology images. Our proposed ensemble image classification method consists of three main parts: the image pre-processing part to adjust the image contrast to show the clear nucleus that can be recognized as the cause of cancer, the image extraction part to extract only important features, and finally the model creation part that generate the model to be used later as an image-based predictor. The experimental results show that the proposed method can predict the colorectal cancer from the colon images with high accuracy.