A Hybrid Metric Transformer-Based Deep Classification Network for Lung Cancer CT Image Classification in Internet of Things

Early diagnosis and effective treatment of lung cancer have always been key focuses in medical research. Deep learning has been widely applied in the auxiliary diagnosis of lung cancer, demonstrating promising application potential and research value. However, the performance of existing deep learning-based lung cancer classification models is often limited by weak inter-class differences and insufficient feature generalization capabilities. To address these challenges, this study proposes a Hybrid Metric Transformer-based Deep Classification Network (HMT-DCN) for lung cancer CT image classification based on the Internet of Things (IoT). The HMT-DCN model integrates the strengths of Vision Transformer (ViT) and Convolutional Neural Networks (CNN) to extract more representative image features, effectively capturing both fine-grained lesion details and contextual information. Additionally, a metric learning strategy is incorporated into HMT-DCN, enabling the model to distinguish between different lung cancer subtypes in the feature space without relying on explicit subtype labels. This enhances the model’s ability to capture subtle inter-class variations. Extensive experiments conducted on real-world lung cancer CT image datasets validate the effectiveness of the HMT-DCN model. The proposed HMT-DCN achieved an accuracy of 0.946, precision of 0.947, and F1-score of 0.949, demonstrating its superiority over existing CNN- and Transformer-based models.