Comparing Artificial Intelligence Approaches to Retrieve Clinical Reports Documenting Implantable Devices Posing MRI...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Data Analysis

Background:

• Identifying patients with implantable devices posing MRI safety risks is crucial for patient care.
• Existing methods for identifying these devices may be time-consuming or incomplete.
• Natural Language Processing (NLP) offers potential for automated identification.

Purpose of the Study:

• To compare the accuracy of expert-derived and ontology-derived NLP approaches for identifying patients with MRI-unsafe implantable devices.
• To determine the prevalence of such devices within various clinical data sources.

Main Methods:

• Retrospective study at an academic hospital comparing two NLP methods: expert-derived and ontology-derived.
• Expert-derived approach used curated terms; ontology-derived approach used Systematized Nomenclature of Medicine-Clinical Terms.
• Analysis of radiology reports, emergency department notes, and other clinical reports.

Main Results:

• Both expert-derived and ontology-derived NLP approaches demonstrated similar accuracy, sensitivity, and specificity.
• The proportion of clinical reports identifying high-risk implantable devices ranged from 1.47% to 1.88%.

Conclusions:

• Artificial intelligence, specifically expert-driven and ontology-driven NLP, are effective and comparable tools for identifying patients with MRI safety-risk implantable devices.
• These NLP methods can be integrated into clinical workflows to enhance patient safety.