This is the authors' abstract. We don't add key points for this paper.
Medical image analysis is critical for disease diagnosis and treatment. However, the complexity and diversity of medical images make effective segmentation without prior information a significant challenge. This paper proposes a Multi-view Clustering Ensemble method via Graph Tensor Learning (MCEGTL) to enhance clustering performance by integrating information from various perspectives. The multi-view learning framework consolidates comprehensive and complementary information from different views, while ensemble learning techniques further improve clustering accuracy. Specifically, a random view grouping strategy captures the interrelationships between multiple views, enabling efficient multi-stage fusion. By expanding the multi-view data into multiple view groups, a consensus bipartite graph is constructed in the early fusion stage by jointly exploring diversity at three levels. Subsequently, diverse base clusterings are obtained through fast graph partitioning of different view groups, which are further fused into a unified bipartite graph for final clustering. The base clustering is refined into approximate graph structures using a theoretical rank constraint with an adaptive weight coefficient. Additionally, the multiple refined connectivity matrices are stacked into a third-order tensor, from which higher-order similarities are extracted via graph tensor learning. This approach also facilitates the propagation of mutual information between different graph matrices. Experiments on several medical image datasets demonstrate the advantages of the proposed MCEGTL compared to state-of-the-art methods.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: