Precise Identification of Gastric Cancer Pathological Differentiation Based on Hyperspectral Imaging and Lightweight Deep Learning Models

Accurate classification of gastric cancer differentiation is crucial for prognosis and treatment decisions. In this study, we propose a lightweight deep learning model-Improved Deep Residual Network (IDRN)-combined with hyperspectral imaging (HSI) to achieve precise identification of gastric cancer tissues. The model incorporates spectral preprocessing, dimensionality reduction, and a residual CNN with attention mechanisms to enhance feature extraction while maintaining efficiency. Comparative experiments with SVM, ResNet50, and ViT models show that IDRN achieves superior performance, particularly in identifying poorly differentiated tissues. Our approach provides a promising tool for computer-aided diagnosis and offers potential for clinical translation.