There is a large, NM-wide effort to improve outcomes of patients with incidental radiographic findings that are inadequately followed up. One key piece to this initiative is in the earliest possible identification of an incidental finding (regardless of whether it will be followed up appropriately). There exists myriad data on previous incidental findings, also, for some findings, there are relatively standard nomenclature applied by radiologists in their free-text reports. Throughout the past year, we collaboratively built and test a “finding detector” that will use a combination of natural language processing and machine learning to identify patients with an incidental radiographic finding. We also worked collaboratively with NM IS another aspect of this project – identifying which findings were followed up appropriately, and which remain “delinquent” in follow up and should be routed to a separate team for review. The results from year one yielded two NLP models that identify lung and adrenal nodules respectively and their respective Epic integrations. Building on this work, we will extend to other types of findings (prostate, for example) and also further enhance the decision support of our existing and future models by providing “highlighted text” back to the clinician to help them hone in on the areas of the radiology report most likely to contain the findings. Finally, if time allows, we will extend this work to the images themselves – attempting to locate incidental findings from the images directly therefore bypassing the report. This could potentially discover truly missed findings.
|Effective start/end date||7/1/20 → 6/30/23|
- Northwestern Memorial HealthCare (Agmt 8/28/20)
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.