Knowledge-assisted Visual Analytics

KnoVA is a research project designed by a team of researchers from the Institute of Software Technology and Interactive Systems at the Vienna University of Technology, aiming at the advancement of the state of the art in Visual Analytics. Visual Analytics is an emerging discipline committed to the development of new methodologies for the analysis of large amounts of data by combining the outstanding perceptual and cognitive capabilities of humans and the enormous information processing possibilities offered by computers. Data analysis is a knowledge-oriented process, to which knowledge constitutes both an input and an output: while prior knowledge is required for interpreting the data as well as for formulating and testing hypotheses. This process generally yields a number of insights, which can be condensed into new knowledge. Nevertheless, prior knowledge is generally tacit: it is used throughout the process, but seldom formalized or made explicit. The working hypothesis of this project is that the Visual Analytics process can be greatly improved by maximizing the use of explicit knowledge throughout each one of its steps, and considering knowledge about the phenomenon described by the data (i.e., domain knowledge) as well knowledge relative to the analytic process itself (i.e., operational knowledge). Such knowledge-assisted Visual Analytics methodologies are generally useful in domains featuring the combination of data and explicit knowledge about specific phenomena. This project will use the clinical domain as application field, where data about patients conditions and treatments is stored under the form electronic health records, and medical knowledge is formally represented in the form of clinical guidelines. In order to achieve the aforementioned results, the project will embrace a human-centric design methodology, through the derivation of an initial set of requirements from an in-depth analysis of users tasks and analytic workflows, and the adoption of an iterative development process driven by evaluation with final users. The idea of designing systems that are able to leverage explicit knowledge for providing a better experience to users in data analysis tasks is not new. Nevertheless, existing approaches have so far focused on the use of the so-called operational knowledge mainly for the adaptation of the visual aspects of the interface. The main scientific innovation of this project lies on the use of explicit domain knowledge, like the Computer-Interpretable Guidelines used in clinical decision-support systems. Moreover, the project aims at enhancing prior approaches to knowledge-assisted analysis by leveraging explicit knowledge at each step, from visual mapping and view manipulation, to the selection and refinement of analytic models, and interaction opportunities.

The KnoVA project investigated the feasibility of a knowledge-assisted approach to Visual Analytics (VA) in the healthcare domain, using uveitis clinical trial data as a primary focus. The research hypothesized that leveraging explicit prior knowledge could improve data analysis, reduce user cognitive load, and increase the number and quality of insights gained. Collaborating with the Medical University of Vienna, KnoVA explored anonymized Patient Report Outcome Measures (PROM) data from a uveitis clinical trial. PROMs are analysis instruments, like textual questionnaires, designed to capture high-dimensional, context-sensitive, and time-oriented quality-of-life data. The results confirmed the original hypothesis and the benefits of using explicit knowledge to enhance the visual data analysis processes. The most significant results centered around developing a knowledge-assisted VA model, utilizing explicit knowledge, guidance-enriched VA, multivariate and temporal network visualization, and applications in medical trials and other domains. The knowledge-assisted VA model incorporates an explicit knowledge source and several knowledge-related processes, which informed the design, development, and evaluation of a VA prototype for exploring and analyzing longitudinal healthcare data. This prototype demonstrated the potential of using explicit knowledge for guidance and steering analysis in crucial moments. The results can be extended to other domains, such as biochemistry, psychology, and education. KnoVA also contributed, more in general, to the advancement of guidance-enhanced VA. The results illustrate how explicit knowledge is exploited to guide the user toward insights, contributing a categorization and a survey of existing guidance approaches, conceptual frameworks for designing effective guidance, and a method for detecting when guidance is needed. KnoVA also explored multivariate and temporal network visualization from various perspectives, including a meta-survey detailing research directions in the field, design and evaluation of network visualization supporting PROMs, and experimental studies on graph structural and temporal encoding. In healthcare, KnoVA's findings were crucial in enhancing the effectiveness and efficiency of data-driven decision-making processes. KnoVA's researchers explored visualization modalities for patients' medical histories, designed and evaluated VA solutions for medical trials using PROMs, and analyzed the user-data-task design principle for digital humanities. In summary, the KnoVA project successfully demonstrated the potential of knowledge-assisted VA in healthcare data, opening new research opportunities in VA and healthcare. Developing new research methods and instruments, such as interactive prototypes and visualization techniques, offers valuable resources for researchers and practitioners to explore, analyze, and communicate complex healthcare data. KnoVA's transdisciplinary approach also incorporated insights from various fields, enabling more holistic and robust solutions to healthcare data analysis challenges.