Author: Curtiland Deville, Rachel B. Ger, Elaina Hales, Heng Li, Todd R. McNutt 👨🔬
Affiliation: Department of Radiation Oncology and Molecular Radiation Sciences, Johns Hopkins University School of Medicine, Johns Hopkins University 🌍
Purpose: Lack of specialized coding knowledge limits the use of data from large SQL repositories. We aimed to develop intuitive user interfaces for a registry that allows users to visualize the relationships between various demographic, prescription, and outcomes data with no specialized knowledge necessary.
Methods: Our center’s proton registry contains all consented patients treated at our proton center with their 3D dose information (e.g., ROIs, DVH), prescription information, demographics, and assessments from on-treatment and after-treatment visits stored in a SQL database. Power BI was used to connect this registry and build user interfaces that model relationships between these datasets. We used the built-in DAX (data analytics expressions) library to connect different tables within our repository and to create statistical calculations. After these connections were made, we added drop down menus to facilitate interaction. Each dashboard was designed such that users could select patient cohorts with drop down menus or by clicking on the visuals themselves.
Results: Three interactive user interfaces were developed and tested for feasibility by physicists and physicians. The first visualizes outcomes data. Users can select disease site, assessment name, grade, and/or date from the dropdown menu to examine outcomes data for the selected subset of patients. The second visualizes prescription and dose information. It allows users to select a region of interest (ROI) to view dose information and overlays all DVHs for the selected patients. The third interface visualizes demographic data. It allows users to see a more complex spread of disease site by ICD10 code and the distributions of age, gender, and race. All interfaces support clicking on the visual itself and using the control bar to select cohorts.
Conclusion: The dashboards can easily display the data within the registry and allow users to quickly and easily identify patient cohorts to perform studies evaluating therapeutic efficacy.