Diagnosis of Lymph Node Metastasis of Pancreatic Cancer Based on Feature Extraction and Information Fusion

The accurate assessment of preoperative lymph node metastasis status in patients with pancreatic cancer is of great significance for formulating reasonable treatment strategies and judging prognosis. However, accurately determining whether lymph node metastasis has occurred before surgery still faces significant challenges. In this paper, a multimodal information fusion method combining two phase CT images (venous phase, arterial phase) and clinical data is proposed, and a deep learning prediction model for lymph node metastasis diagnosis of pancreatic cancer is constructed. The model first uses a dual channel feature extraction network to extract deep level imaging features from two-phase CT images; Subsequently, the extracted image features were effectively fused with the patient's clinical features using an attention based information fusion method. Based on the data set used in this study, the experimental results show that the proposed AI model performs better in classification performance than traditional imageomics methods, providing a feasible and effective means for accurate prediction of preoperative lymph node metastasis of pancreatic cancer.