Interactive Visual Exploration of Causal Structures for Neuropathic Pain Diagnosis

Abstract: Revealing causal structures from observational data is an essential task in many data analysis issues across various domains, such as natural sciences, business, and healthcare. In healthcare, neuropathic pain is one of the most common medical problems, whose diagnosis process has well-understood causal structures. Causal structures are commonly visualized as a directed acyclic graph (DAG) or a node-link diagram, in which nodes represent variables, and edges represent causal relationships between data dimensions. However, these simple static graphs do not convey sufficient information for either an intuitive interpretation for novel viewers, or an in-depth exploration for experts. In this study, the visualization of causal structures for neuropathic pain diagnosis is set into context. An interactive system that integrates application-specific visualization, i.e. a discomfort drawing and a spinal cord diagram, into causality visualization is developed. It is further evaluated by a domain expert on neuropathic pain and a researcher in causal discovery through semi-structured interviews. The results show that the system reveals the causal structures for neuropathic pain diagnosis in a more intuitive, efficient way, and conveys more focused information compared to traditional node-link diagrams. The system is also demonstrated to be helpful to the medical community in neuropathic pain diagnosis, not only for doctors but also for patients.