Classification of Lung Cancer Using Ensemble Deep Neural Network Through Computed Tomography Images

Lung cancer is recognized as a leading cause of cancer-related mortality, underscoring the urgent need for accurate diagnostic tools to improve early detection. This study addresses the significant challenges of low accuracy and limited effectiveness in lung nodule detection, which can hinder timely diagnosis. Combining DenseNet 201 and ResNet 50, an ensemble deep neural network model is proposed to enhance classification performance. A comparative analysis is conducted to evaluate the effectiveness of the ensemble model using both original CT images and segmented outputs generated by a U-Net model. It is found that the ensemble model significantly outperforms when fed with original CT images, achieving an impressive 100% accuracy in both training and validation phases, along with a 100 % Area Under the Curve (AUC) score. The proposed model reported 16.26 MFLOPs in terms of computational resources. The findings highlight the importance of utilizing original CT images to improve classification reliability and diagnostic capabilities. The proposed approach improves the precision of lung cancer binary classification of malignant versus normal CT scan images.