ViMDT: a clinical data visual analysis tool for multi-disciplinary treatment of lung cancer.

Multi-disciplinary treatment (MDT) has become a routine practice in clinical cancer diagnosis and treatment, playing an indispensable role in clinical decision-making. By integrating expertise from multiple disciplines, MDT provides patients with individualized diagnosis and treatment strategies. However, there is not yet a specialized clinical data visualization tool for MDT. This paper develops a novel clinical data analysis visualization tool for MDT, which analyzes in-depth and displays patient data comprehensively. Specifically, this tool designs a latent Dirichlet allocation (LDA)-based visualization model for clinical unstructured data, and Z-Score-3σ transformation and hierarchical strategies for clinical structural data. Moreover, we propose to predict personalized anti-tumor drug efficacy based on topic keywords. The results showed that, compared with users who did not use the tool, the time cost in MDT decision-making for users who used the tool was reduced by 26.17%. Furthermore, the proposed drug efficacy prediction method achieved an accuracy rate of 71.08% on a dataset of 958 patients with non-small cell cancer treated with anti-tumor drugs. The proposed tool is potentially helpful for doctors in MDT tasks by vividly visualizing the large-scale complex clinical data and improving the MDT efficiency.