Deep Learning–Based Lung Cancer Classification Using CNN, VGG16, an‎d ResNet50 Architectures

5/10/2026 12:00:00 AM

The significance of the pro‎mp‎t lies in the emphasis placed on the need for the timely an‎d accurate diagnosis of pulmonary diseases to improve patient health an‎d lower mortality rates. It also focuses on the fundamental role that various medical imaging modalities play in the diagnosis of various health conditions. The interpretation of the modality by the radiologist, however, is a time-consuming process that carries a high probability of error. The recent advances in deep learning methods, however, show promising results an‎d potential for the automation of the interpretation of the modality. Study purpose an‎d objectives: The objectives of the research include the presentation of a new methodology for the classification of lung diseases using convolutional neural networks (CNN) an‎d other related deep learning methods in the interpretation of the modality. The methodology for the classification of the disease in the modality is divided into three phases. The implementation of the methodology, however, relies on the proposed CNN model, an‎d the results are obtained from the implementation of the model in the context of three different architectures of the CNN model, including the simple CNN, VGG16, an‎d ResNet50. The results obtained from the implementation of the different architectures of the model, however, are presented in the results section. Moreover, the eva‎luation of the performance of the model relies on traditional performance eva‎luation metrics, including accuracy, precision, recall, F1 score, confusion matrix, ROC analysis, an‎d AUC values. The results obtained from the implementation of the model, however, show that the transfer learning model performs better than the custom CNN model. Moreover, the best results in the context of classification accuracy an‎d AUC values are obtained from the ResNet50 model. To assess the results in the context of real-world scenarios, the results in the context of AUC values are compared to the results obtained from the traditional radiologists. The results, therefore, validate the effectiveness an‎d appropriateness of the CNN model in the extraction of discriminative features from the pulmonary images. Moreover, the results obtained from the implementation of the model in the context of the proposed framework show the potential an‎d effectiveness of the model in the automated diagnosis of the disease.

شبكھھاي عصبي كانولوشني , طبقھبندي بيماريھاي ريوي , تحليل تصاوير پزشكي

Convolutional Neural Networks, , Lung Disease Classification , Medical Image Analysis

Abdulrahman Al‑Dulaimi

Dr. Behrouz Minaei