Multimodal Siamese Model for Breast Cancer Survival Prediction

Breast cancer is one of the most common cancers, and accurate survival prediction can help doctors make appropriate treatment decisions for patients, thereby improving patient survival rates. To enhance the performance of survival prediction for breast cancer patients, a multimodal siamese model is proposed. The model utilizes a Siamese-RegNet to extract representative features, which not only considers the relationship between pathological patches but also extracts survival-related features. Moreover, The proposed model takes both pathological images and clinical data as multimodal input, which is essential to survival prediction and can improve predictive performance. The experimental results on the public dataset TCGA-BRCA and external test dataset GMUCH-BRCA demonstrate that the proposed model achieves a better concordance index (C-index) of 0.756 and 0.567, respectively.