Deep Learning-Based Multimodal Integrated Risk Stratification Prediction Model for Non-Small Cell Lung Cancer

Lung cancer is the most common type of cancer and has a high mortality rate. With the development of precision medicine, the treatment strategy of advanced non-small cell lung cancer (NSCLC) has shifted to individualized targeted therapy and immunotherapy, but the 5-year survival rate is still low. Traditional survival risk prediction methods are based on clinicopathological risk factors, but cannot fully reflect tumor biology. Therefore, researchers have begun to focus on biomarkers, among which tumor mutation burden (TMB) is an important biomarker for immunotherapy. In this study, a deep learning model was constructed to predict the risk stratification of TMB markers using National Cancer Institute's Clinical Proteomic Tumor Analysis Consortium (CPTAC) data, histopathological images and clinical information of NSCLC patients. The model contains two branches for processing histopathological images and clinical information, and is deeply fused by a bilinear pooling layer. Experimental results show that the multimodal model can comprehensively utilize two different types of data to improve the accuracy and reliability of patient risk prediction, which helps to guide patients to choose appropriate treatment strategies.