Multi-Class Classification of Lung Cancer Using Machine Learning and Image Processing

In this research paper, we present lung cancer classification using machine learning techniques. We used several techniques to classify the CT-scan images into three classes which are normal, benign and malignant (cancerous) classes. The dataset was obtained from a reliable source having 1096 images overall. The dataset was preprocessed to have balanced classes and feature extracted using multiple image filters. Many experimental cases were conducted using various machine learning techniques which are decision tree, random forest, support vector machine and K-nearest neighbor, and evaluated considering binary class classification and multi-class classification, which are normal vs. non-normal, cancerous vs. non-cancerous, and normal vs. benign vs. cancerous. Five performance metrices which are confusion matrix, accuracy, recall, precision and F1-score were considered. The results show that the random forest and K-nearest neighbor techniques that we consider performed very well. The normal class (healthy subjects) and the benign (not-harmful tumor) class can be correctly classified 97%-99%. Overall, we obtained the accuracy of 99.73 % for 2-class classification, and accuracy of 98.69 % for the 3-class classification.