DeepPanNet – A Multi-Layered Graph Neural Networks (ML-GNN) for Tumor Detection and Classification

In medical imaging, Pancreatic tumor detection, and classification is challenging to detect accurately. This paper introduces a DeepPanNet, the Multi-Layered Graph Neural Network (ML-GNN) system, to detect and classify tumors. DeepPanNet is a combination of graph-based representations allowing the capture of complex spatial and relational dependencies in medical imaging data. The ML-GNN consists of graph convolutional layers, attention modules, and multi-scale feature aggregation; this architecture improves classification accuracy without losing structural information. The pre-trained model VGG19 that traines the model on pancreatic tumor images. The model performance on publicly available pancreatic tumor datasets indicates better performance over conventional Deep learning (DL) models. Experimental results show that DeepPanNet outperforms in terms of performance metrics, proving its effectiveness for real-world clinical settings and applications. Moreover, attention-based feature selection improves the interpretability of the model and helps a medical professional in decision-making.