Lung cancer detection from CT scans using modified DenseNet with feature selection methods and ML classifiers

Abstract

Lung cancer is a highly life-threatening disease worldwide, and detection is crucial. In this study, the Kaggle chest CT-scan images dataset was used to identify lung cancer in four categories: adenocarcinoma, large cell carcinoma, squamous cell carcinoma, and normal cell. A unique Deep Learning (DL) based method was suggested by modifying the DenseNet201 model and adding layers to the original DenseNet framework to identify lung cancer disease. Two feature selection methods were used to select the best features extracted from DenseNet201, which were then applied to various ML classifiers. The system's performance was evaluated using a confusion matrix, ROC curve, Cohen's Matthews Correlation Coefficient (MCC), Kappa score (KS), 5-fold method, and p-value. The proposed system achieved a high accuracy of 100 perent, an average accuracy of 95 percent, and a p-value of less than 0.001 after applying a 5-fold method. This study highlights the potential of using computer technology and ML methods to improve the accuracy of a lung cancer diagnosis from CT scans.