1 / 22

Computational Diagnostics: A Novel Approach to Viewing Medical Data

This paper presents a computational diagnostic tool that combines interactive visualizations to view medical data, specifically focused on acute lymphoblastic leukemia patients. The tool includes matrix visualization, parallel coordinate visualization, and user interactions to analyze patient variables and identify trends and patterns in the data.

cvidrio
Download Presentation

Computational Diagnostics: A Novel Approach to Viewing Medical Data

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. July 2nd, 2007 Computational Diagnostics: A Novel Approach to Viewing Medical Data 5th International Conference on Coordinated & Multiple Views in Exploratory Visualization Dr. Ketan K. Mane Los Alamos National Laboratory, Los Alamos, New Mexico kmane@lanl.gov Dr. Katy Börner School of Library and Information Science, Indiana University, Bloomington katy@indiana.edu Presented by: Bruce Herr

  2. Outline • Motivation • Dataset – Acute Lymphoblastic Leukemia • Computational Diagnostic Tool • Conclusion

  3. Motivation • Electronic Health Record (EHR) Initiative – digitally store patient medical records Challenge – 1) Combine data from different sources 2) Make sense of data for analysis 3) Acquire implicit knowledge • Existing EHR include - • Reproduce ‘electronic view’ of a paper record – for data entry • Display patient medical history in graphical time-series format • Add 2D/3D images to patient profile, link images for interaction 2D / 3D linked images LifeLines showing time-series information

  4. Computational Diagnostics - Dataset Details Diagnostic data variables from medical records for Acute Lymphoblastic Leukemia (ALL) patients are categorized into: a. OutcomePatient Variables: relapse, relapse site, alive/death status, and LDKA b. BiologyPatient Variables: immunophenotype, genetic condition, WBC, Hgb, platelets, • and CNS • c. HostPatient Variables: diagnostic age (ageDx), gender, and race • d. Treatment • Patient Variables: BM 7 and BM 14 e. Social Factors • Patient Variables: MFI-class, education level, %single family members, and % family employment • All data was provided by Dr. Susanne Ragg, Julie Haydon and Jada Pane at IUPUI

  5. Computational Diagnostics – Tool Coupling of different interactive visualizations such as Matrix visualization phenotype, prognosis and combined view. Parallel Coordinate Visualization Coupled Windows

  6. Computational Diagnostics – Interactive Visualization System Architecture

  7. Matrix Visualization – Phenotype View • Data is shown independent of other variables. • Color codes help to provide a quick insight into patient medical condition.

  8. Matrix Visualization – Prognosis View • Color codes indicate event free survival in percent (%EFS). • All variable values are dependent on other variable values.

  9. Matrix Visualization – Prognosis View • Color codes indicate event free survival in percent (%EFS). • All variable values are dependent on other variable values.

  10. Matrix Visualization – Prognosis View • Color codes indicate event free survival in percent (%EFS). • All variable values are dependent on other variable values.

  11. Matrix Visualization – Combined View • Facilitates selection of phenotype/prognosis view for individual diagnostic variables.

  12. Parallel Coordinates Visualization • Uses one axis for each data variable. • For each patient, all data values on different parallel axis are connected. • All patient graphs are shown here. Single or multiple patients can be selected and studied in detail.

  13. Parallel Coordinates Visualization Tool-tip display to show diagnostic values of selected patient

  14. Parallel Coordinates Visualization – User Interactions Display axes-labels to mark different regions/values along axes • Numerical landmarks along axes showing values for quantitative variables. • Category labels used along axes show values for nominal variables.

  15. Parallel Coordinates Visualization – User Interactions Display zones to show severity values for different variables • Triangular zones indicate variables with quantitative values. • Rectangular zones are used for variables with nominal values.

  16. Parallel Coordinates Visualization – User Interactions Axis selection to study global variations in patient values • Single axis can be selected to study the trend in patient values. • Red-to-green gradient used to indicate values along the selected axis. [Red = High value, Green = Low value]

  17. Parallel Coordinates Visualization A subset of patents can be selected and examined as a group

  18. Parallel Coordinates Visualization Simultaneous display of patient groups to study differences Patient Group 1 Patient Group 2 Patient Group 1 & 2

  19. Parallel Coordinates Visualization Multiple Coordinated Views • Patient can be selected and color coded in matrix view. • Corresponding patient lines are highlighted in parallel coordinate view.

  20. Conclusion • Computational Diagnostic Tool provides – • Quick overview of patients medical condition • Color visual cues (in matrix view) provide global dataset overview and help to - • Identify patients with worst values for different variables • Compare patients to identify overlaps and differences in values • Line patterns (in parallel coordinates view ) show data trends and help to – • Reveal patient profile • Compare patients profile within a group or among groups • Multiple coordinated view – both view complement each other • Matrix view – helps identify patterns and worse case conditions • Parallel coordinates – helps identify patient groups and compare trends

  21. Acknowledgements • All the work was done when at Indiana University, Bloomington • Thanks to Dr. Susan Ragg for her insights on the Acute Lymphoblastic Leukemia dataset and also to Julie Hayden and Jada Pane for making the ALL dataset available. • This work was supported by • National Science Foundation under DUE-0333623 • 21st Century Grant for the Center of Excellence for Computational Diagnostics

  22. Thank You !!! For additional details email: Ketan K. Mane kmane@lanl.gov

More Related