Interactive Visual Analysis and Causal Modeling on Clinical Data

Juliane Müller is a Research Associate at the Department of Neurology, Otto-von-Guericke University Magdeburg, Germany. Her main area of research is in visually supporting hypotheses generation for biomarker identification as well as in assisting in causal model creation and comprehension.

Abstract:
The enormous amount of medical data recorded in clinical routine allows for new facilities in hypotheses generation and identification as well as automated causal model generation for Clinical Decision Support (CDS). Although there is always a validation step by domain experts required before taking generated causal relations for granted, automatically-generated causal models can serve as an initial draft for disease-specific causal model creation and validation. To improve the correctness of these drafts, variables of interest first need to be identified, since fewer involved variables in model generation lead to better results among the generated causal relations. When validating and applying a causal model in clinical routine, comprehension of the underlying reasoning is a crucial prerequisite, especially in the medical domain.
In this talk, I will present a pipeline for generating causal models from data. Here, I’m especially focusing on the visual assistance in identifying variables of interest and on the comprehension for validation and usage in daily clinical routine.