Entity Extraction for Clinical Notes, a Comparison Between MetaMap and Amazon Comprehend Medical

Area of Science:

• Clinical Informatics
• Natural Language Processing in Healthcare
• Biomedical Data Extraction

Background:

• Clinical notes, such as discharge summaries, are vital sources of patient information but are often unstructured or semi-structured.
• Extracting diseases, risk factors, and treatments from these notes is crucial for improving healthcare quality and facilitating clinical research.
• Making this extracted information accessible to computerized systems requires robust automated entity extraction methods.

Purpose of the Study:

• To compare the automated medical entity extraction capabilities of two distinct tools: MetaMap (MM) and Amazon Comprehend Medical (ACM).
• To evaluate the performance of MM and ACM using standard metrics for information extraction.
• To determine which tool offers superior accuracy in identifying and extracting medical entities from clinical text.

Main Methods:

• Two automated entity extraction tools, MetaMap (MM) and Amazon Comprehend Medical (ACM), were utilized.
• Performance evaluation was conducted using recall, precision, and F-score metrics.
• Both tools processed a dataset of clinical notes to extract medical entities.

Main Results:

• Amazon Comprehend Medical (ACM) demonstrated higher average recall compared to MetaMap (MM).
• ACM achieved a greater average precision than MM in medical entity extraction.
• The overall F-score for ACM was superior to that of MM, indicating better overall performance.

Conclusions:

• Amazon Comprehend Medical (ACM) is a more effective tool for automated medical entity extraction from clinical notes than MetaMap (MM).
• The findings suggest ACM's suitability for applications requiring accurate and efficient processing of clinical text data.
• Higher performance metrics for ACM indicate its potential to enhance data accessibility for healthcare quality and research.