Interactive Natural Language Processing for Clinical Text

[摘要] Free-text allows clinicians to capture rich information about patients in narratives and first-person stories. Care providers are likely to continue using free-text in Electronic Medical Records (EMRs) for the foreseeable future due to convenience and utility offered. However, this complicates information extraction tasks for big-data applications. Despite advances in Natural Language Processing (NLP) techniques, building models on clinical text is often expensive and time-consuming. Current approaches require a long collaboration between clinicians and data-scientists. Clinicians provide annotations and training data, while data-scientists build the models. With the current approaches, the domain experts - clinicians and clinical researchers - do not have provisions to inspect these models or give direct feedback. This forms a barrier to NLP adoption and limits its power and utility for real-world clinical applications.Interactive learning systems may allow clinicians without machine learning experience to build NLP models on their own. Interactive methods are particularly attractive for clinical text due to the diversity of tasks that need customized training data. Interactivity could enable end-users (clinicians) to review model outputs and provide feedback for model revisions within an closed feedback loop. This approach may make it feasible to extract understanding from unstructured text in patient records; classifying documents against clinical concepts, summarizing records and other sophisticated NLP tasks while reducing the need for prior annotations and training data upfront.In my dissertation, I demonstrate this approach by building and evaluating prototype systems for both clinical care and research applications. I built NLPReViz as an interactive tool for clinicians to train and build binary NLP models on their own for retrospective review of colonoscopy procedure notes. Next, I extended this effort to design an intelligent signout tool to identify incidental findings in a clinical care setting. I followed a two-step evaluation with clinicians as study participants: a usability evaluation to demonstrate feasibility and overall usefulness of the tool, followed by an empirical evaluation to evaluate model correctness and utility. Lessons learned from the development and evaluation of these prototypes will provide insight into the generalized design of interactive NLP systems for wider clinical applications.

[发布机构] the University of Pittsburgh