Saja, Al-alawneh
(2021)
Clinical Text Analysis using Visual Analytics for Cancer Patient Cohort Identification.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Due to the complexity nature of cancer patients’ records and clinical notes, extracting and summarizing the required data to identify a cohort of interest is a challenge for cancer researchers. DeepPhe pipeline was developed to support cancer cohort identification and hypothesis generation by extracting deep phenotypes from cancer patient’ electronic health records using natural language processing, rule-based and machine learning techniques. The pipeline generated high-level summaries from individual mentions as phenotypes and visualized them using a web-based visual analytics interface DeepPhe-Viz to create a longitude representation of cancer process.
In this study, we extended the functionalities of DeepPhe-Viz interface by combining the extracted data from the NLP pipeline and visual analytics to empower researcher’s ability to mine and uncover new and challenging insights about cancer population in the textual documents of the EMR data.
To advance the capabilities of the DeepPhe-Viz, first, we implemented an interactive heatmap visualization that viewed high-level representation of all the extracted terms from clinical documents. This feature enabled cancer researchers to investigate the rich contents of the full clinical document text to identify additional key variables such treatment that are extracted using the DeepPhe pipeline. Second, we implemented an interactive Sankey diagram visualization to aggregate all the transitions of predefined episodes of care for a cohort of cancer patient which represent temporal events in the clinical documents. This feature is essential for gaining a deeper understanding of different patterns and trends of cancer treatment mentioned in EMR.
Finally, we evaluated the usability of DeepPhe-Viz interface to identify a cohort of patients and to drill down to more details about patients. User studies results including qualitive and quantitative feedback indicated the usefulness and feasibility of the DeepPhe-Viz interface to the cancer investigators to conduct cancer retrospective studies using EMR data
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
29 October 2021 |
| Date Type: |
Publication |
| Defense Date: |
22 July 2021 |
| Approval Date: |
29 October 2021 |
| Submission Date: |
17 September 2021 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
147 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Medicine > Biomedical Informatics |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Natural language processing, cohort identification, visual analytics |
| Date Deposited: |
29 Oct 2021 13:19 |
| Last Modified: |
16 Nov 2021 15:12 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/41693 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}