Deep Learning Based Cancer Survival Analysis Algorithm

Abstract

Histopathological image examination is the gold-standard tool of cancer diagnosis. However, it is subjective to only rely on doctor qualitative analysis of pathological image. What’s more, it is easy to miss pathological information. The texture of pathological images is complex and highly heterogeneous, so it is difficult to extract the deep features of histopathological images. In this paper, a deep learning-based cancer survival analysis algorithm is proposed to process the histopathological image data, which can effectively extract the histopathological image features and improve the accuracy of the model. The average C-Index of the proposed method is 0.7233, which is improved by 28.0% and 19.9% compared with two state-of-the-art methods.