MVHRL: Multi-view Hierarchical Representation Learning for Subtype Classification of Non-small Cell Lung Cancer

Accurate classification of histological subtypes in non-small cell lung cancer (NSCLC) through the analysis of computed tomography (CT) images is essential for clinician diagnosis. Although previous multi-view methods have made significant progress, they suffer from the conflict between the consistency learning task and the reconstruction task when striving to learn discriminative view-common representations. In this paper, we propose a novel multi-view hierarchical representation learning method (MVHRL) for accurate NSCLC histological subtype classification. MVHRL introduces a hierarchical representation learning network to generate low-level representations and high-level representations for each view to accomplish the reconstruction task and the consistency learning task, respectively, thereby alleviating the conflict. Comprehensive experimental results validate the effectiveness of our method in resolving the conflict issue, and fully mine the view-common information in enhancing the performance of the model.