Integrative Analysis of Multi-view Histopathological Image Features for the Diagnosis of Lung Cancer

Lung cancer is one of the most widely spread cancers in the world. So far, the histopathological image remains the gold standard in diagnosing lung cancers, and multiple types of pathological images features have been associated with lung cancer diagnosis and progression. However, most of the existing studies only utilized single type of image features, which did not take advantages of multiple types of image features. In this paper, we propose a Block based Multi-View Graph Convolutional Network (i.e., BMVGCN), which integrates multiple types of image features from histopathological images for lung cancer diagnosis. Specifically, our method utilizes the block-based bilinear combination model to fuse different types of features. By considering the correlation among different samples, we also introduce the Graph Convolutional Network to exploit the correlations among samples that could lead to better diagnosis performance. To evaluate the effectiveness of the proposed method, we conduct the experiments for the classification of the cancer tissue and non-cancer tissue in both Lung Adenocarcinoma (i.e., LUAD) and Lung Squamous Cell Carcinoma (i.e.,LUSC), and the discrimination between LUAD and LUSC. The results show that our method can achieve superior classification performance than the comparing methods.

Automated summary

In this paper, a Block-based Multi-View Graph Convolutional Network (BMVGCN) is proposed for lung cancer diagnosis, which integrates multiple types of image features to improve diagnostic performance. Experimental results demonstrate that our method outperforms other methods in lung cancer classification tasks.