Optimizing Lung Cancer Detection with Support Vector Machines and Random Forest Classifiers

Lung cancer is one of the most frequent causes of cancer deaths around the world and early diagnosis is vital to extend patients' lives. A novel automatic lung cancer detection technique utilizing SVM and RF classifiers is presented. The data is sets of computed tomography (CT) scan images of a patient that after image normalization and noise removal are separated from possible tumours regions. Feature extraction is applied for extracting more relevant characteristics from the segmented images, which may include texture, shape and intensity features Both SVM and Random Forest classifiers classes the images into cancerous and non cancerous images. Furthermore, the performance of these models is contrasted with other machine learning models like k-Nearest Neighbours (KNN) and Logistic Regression (LR). Performance-wise, the random forest classifier is superior to SVM and other machine learning algorithms; overall accuracy and sensitivity were established at 91% as opposed to 88.5% by SVM. The paper proves that applications of machine learning, Random Forest in particularly, can support the early detection of lung cancer which may lead to the enhancements of clinical diagnosis.