Comparative Analysis of Naïve Bayes and Support Vector Machine Algorithms for Mediastinal and Lung Cancer Classification

Abstract

Small Cell Lung Cancer (SCLC) is a lung cancer often found in the mediastinum and hilum of the lung, involving the lymph nodes. This condition often makes it difficult to differentiate between SCLC lung cancer and mediastinal lymphoma through radiological examinations, resulting in some cases showing SCLC lung cancer being misdiagnosed as mediastinal lymphoma. This study aims to create a classification model for both cancers based on digital image processing using 180 images of SCLC lung cancer and 180 images of mediastinal lymphoma cancer. The preprocessing stage includes a median filter and CLAHE, segmentation using Otsu thresholding, first-order and second-order statistical feature extraction and feature selection from both orders. Using grid search optimisation, classification was performed using Naïve Bayes and Support Vector Machine. The results showed the highest accuracy in Naïve Bayes with an accuracy of 98.61%, while Support Vector Machine produced a testing accuracy of 99.16%.