Maurine Tong, Presenter: Nothing to Disclose

William Hsu PhD, Abstract Co-Author: Nothing to Disclose

Rick K. Taira PhD, Abstract Co-Author: Nothing to Disclose

Numerical information presented in figures, tables, and text within a clinical trial report is the evidence that guides clinical decision making and evidence-based medicine (EBM). However, hurdles exist when attempting to extract and apply this information. The published report has data scattered throughout the report and the context behind each piece of data is not automatically extracted. In addition, the clinical trial study describes specific experiments performed on a pre-defined patient population and comparisons between reports to determine subtle differences in trial design requires critical review by a domain expert. The goal of this research is to demonstrate a system with a context-driven representation to: 1) synthesize fragments of information found in clinical trial reports, 2) assess the strength and quality of a the study, 3) compare knowledge from similar trials. Previously, we have demonstrated a framework to organize numerical data within one clinical trial report. In this exhibit, we further develop this system to include the ability to retrieve and compare similar trials based on similar population characteristics, methods, interventions, or survival metrics.

We demonstrate the tool using clinical trial papers in the domain of non-small cell lung cancer (NSCLC). The usability and satisfaction of the system was evaluated using a 10-point Likert scale. The query outputs from the system were evaluated by domain experts for scientific significance.

A structured representation was demonstrated with an interactive visualization. The visualization performed common queries, aided in interpretation and has implications for furthering scientific discovery.

In this exhibit, we have specified a representation for the purpose of synthesizing and integrating selected information from clinical trial reports in the domain of NSCLC. The tool transforms free-text reports to the target representation with functions allowing comparisons between trials. This research is relevant to radiology researchers involved in comparing and synthesizing clinical trial results and provides the basis for inductive reasoning using evidence from trial studies.

Tong, M, Hsu, W, Taira, R, A Knowledge-based Representation of Clinical Trial Reports for Evidence-based Decision Support.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14016310.html