Automated abstraction of myocardial perfusion imaging reports using natural language processing

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence

Background:

• Myocardial perfusion imaging (MPI) findings are in free-text reports, crucial for research.
• Manual review of these reports is time-consuming and impractical for large datasets.
• Automated abstraction methods are needed to efficiently extract MPI results for research.

Purpose of the Study:

• To develop and validate a natural language processing (NLP) algorithm for abstracting MPI reports.
• To assess the accuracy and efficiency of the NLP algorithm in extracting MPI results.
• To analyze patient outcomes based on abstracted MPI data.

Main Methods:

• Developed an NLP algorithm to automatically abstract MPI reports.
• Validated the NLP algorithm against double-blinded cardiologist reviews.
• Conducted secondary analyses on 16,957 MPI tests from patients with suspected acute coronary syndrome (ACS).

Main Results:

• The NLP algorithm demonstrated high sensitivity (96.7%) and specificity (98.9%) for MPI categorical results.
• Agreement between the NLP algorithm and physician reviewers was comparable.
• Abnormal MPI results correlated with higher rates of 30-day acute myocardial infarction or death.
• Report quality issues were identified, impacting both communication and automated abstraction.

Conclusions:

• NLP offers an accurate and efficient approach for abstracting results from free-text MPI reports.
• This automated method can facilitate future research on the benefits of MPI studies.
• Further validation in diverse clinical settings is recommended.