Feature: Making the case for natural language processing
 |
Daniel L. Rubin, an assistant professor in the department of radiology at Stanford University in Palo Alto, Calif., noted that NLP has many faces, based on the desired task, including:
- Text Classification: A text report or sentences in the report are run through a computer classifier program to label the reports, such as for automated ICD coding.
- Name Entity Recognition: Recognizes findings, diseases, devices and diagnoses in reports.
- Information Extraction: Pulls out from reports particular types of factual statements (such as the anatomic location of an imaging finding) or recommendations.
- Information Retrieval: Searches a large database of text reports for those that match certain query criteria.
One of the tools NLP might assist business intelligence with is name entity recognition, according to Rubin. “People will want to look at practices of ordering, ICD-9 coding and more to figure out if particular imaging findings are positive or negative.” However, use cases need to be pieced together so that NLP ultimately proves to support best practices among evidence-based processes.
Fueling a need for NLP in business intelligence tools, NLP’s text classification and information extraction capabilities will provide a good use case, according to Rubin, who refers to radiation exposure concerns. “It will be vital to track and monitor radiation dose by exam type and its timestamp,” Rubin mentioned.
Another use case will be looking toward imaging appropriateness and use. Questions like “Are we doing too much of a certain exam?” might be answered using NLP’s text-mining capabilities to recognize findings as positive or negative.
Unfortunately, Rubin said that current methods of NLP may not be ready for primetime. “Detecting negation is critical and a fundamental level of NLP,” said Rubin. “You have to validate any NLP system in a group of use cases. We haven’t seen NLP largely take off because there isn’t enough work done yet to validate NLP for all the needed use cases.”
However, Rubin does believe that as use cases accumulate, NLP’s true value will emerge.