A lightweight CNN for enhanced non-small cell lung cancer classification using CT scan image.
Authors
Affiliations (7)
Affiliations (7)
- Faculty of Information Technology, Beijing University of Technology, Beijing, 100124, China.
- Department of Computer Science, University of the Punjab, Lahore, 54590, Pakistan.
- Beijing Engineering Research Center for IoT Software and Systems, Beijing, 100124, China.
- Shandong Research Institute of Industrial Technology, Jinan, 250000, China.
- Artificial Intelligence and Data Analytics (AIDA) Lab, CCIS, Prince Sultan University, 11586, Riyadh, Saudi Arabia. [email protected].
- School of Systems and Technology, Department of Artificial Intelligence, University of Management and Technology, Lahore, 54000, Pakistan. [email protected].
- Artificial Intelligence and Data Analytics (AIDA) Lab, CCIS, Prince Sultan University, 11586, Riyadh, Saudi Arabia.
Abstract
Lung cancer is a leading cause of cancer-related mortality worldwide, and its early and accurate detection is critical for improving patient outcomes. Computed tomography (CT) scans are widely used to diagnose lung cancer; however, the accuracy of diagnosis often depends on the expertise of radiologists. Recently, deep learning-based clinical decision support systems have shown promise in assisting diagnosis by providing reliable and consistent predictions. In this paper, we propose MiniConvNet, a lightweight convolutional neural network designed to detect and classify non-small cell lung cancer (NSCLC) and its subtypes, adenocarcinoma, squamous cell carcinoma, and large cell carcinoma, from CT images. We also evaluate the model's generalizability on a histopathological lung cancer dataset, demonstrating its robustness across imaging modalities. We benchmark MiniConvNet against several established CNN architectures, including ResNet50, VGG16, VGG19, Inception V3, MobileNetV3Small, EfficientNetV2B0, and ConvNeXtTiny, under identical experimental conditions. Extensive experiments on two publicly available datasets show that MiniConvNet achieves competitive or superior performance compared to the baselines while maintaining a significantly smaller model size and faster inference. These results highlight MiniConvNet's potential as an efficient and deployable tool for lung cancer subtype classification in resource-constrained clinical settings.