Using Natural Language Processing of Free-Text Radiology Reports to Identify Type 1 Modic Endplate Changes

Area of Science:

• Radiology and Medical Imaging
• Natural Language Processing
• Clinical Informatics

Background:

• Electronic Medical Record (EMR) systems offer access to radiology reports, crucial for quality improvement and research.
• Natural Language Processing (NLP) is essential for converting unstructured EMR text into usable data.
• Identifying Type 1 Modic endplate changes is important for clinical trials and targeted interventions.

Purpose of the Study:

• To assess the feasibility of using NLP to identify patients with Type 1 Modic endplate changes from spine MRI reports.
• To develop and evaluate a rule-based NLP algorithm for this identification task.

Main Methods:

• A dataset of 458 lumbar spine MRI reports was randomly selected.
• Four annotators identified reports containing Type 1 Modic changes.
• A rule-based NLP algorithm using regular expressions was implemented in Java.
• Performance metrics including recall, specificity, precision, and F1-score were calculated.

Main Results:

• The prevalence of Type 1 Modic change in the dataset was 10%.
• The NLP algorithm achieved a recall of 0.70 and a specificity of 0.99.
• Precision was 0.90, negative predictive value was 0.96, and F1-score was 0.79.
• Specificity was higher than recall due to reporting variability and effective negation algorithms.

Conclusions:

• A rule-based NLP approach is effective for identifying patients with Type 1 Modic changes in spine MRI reports.
• The algorithm demonstrates high efficacy in identifying relevant cases with minimal false positives.
• This method supports quality improvement and clinical research by efficiently extracting valuable data from EMRs.