Abstract—Lung cancer is a lethal ailment which has a significant fatality rate among individuals affected by the disease. Timely detection and accurate staging of lung cancer can significantly improve patient survival rate. Computed Tomography (CT) scans are usually employed for diagnosing lung cancer, but manual examination can be slow and error-prone. To address this issue, deep learning techniques are being utilized to speed up and improve the accuracy of detecting cancerous and non-cancerous CT scans. Therefore, this study introduced an innovative transfer learning method aimed at improving the precision of lung cancer classification. The proposed method was built based on the EfficientNet model, modified with additional custom Convolutional Neural Network (CNN) layers and an attention mechanism for accurate lung cancer classification. Experimental analysis was conducted, utilizing eight variants of the modified EfficientNet (B0–B7) using three lung cancer CT scan datasets, comprising IQ-OTH/NCCD, Chest-CT scan, and LIDC-IDRI, grouped into 3, 4, and 2 classes respectively. Various data augmentation techniques were utilized to address the problem of class imbalance and mitigate any biases present. The model achieved accuracies of 99.5%, 98.0%, and 90.3% on the IQ-OTH/NCCD, Chest-CT scan, and LIDC-IDRI datasets, respectively. The results depict that the modified EfficientNetB1 performed better than other presented approaches with respect to both accuracy, sensitivity, F1-Score, and precision. The outcome also indicates that the presented method is more appropriate for multi-class classification of lung cancer.
 
Keywords—transfer learning, EfficientNet, Computed Tomography (CT), lungs cancer, classification, health risks


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).