Building and Evaluating an Orthodontic Natural Language Processing Model for Automated Clinical Note Information...

Area of Science:

• Biomedical Informatics
• Dental Research
• Machine Learning in Healthcare

Background:

• Malocclusion diagnosis and treatment planning face challenges due to subjective assessments and unstructured data in electronic dental records (EDRs).
• Extracting valuable clinical information from free text in EDRs is complex, hindering consistent and objective orthodontic care.
• This study addresses the need for automated data extraction to improve orthodontic treatment planning.

Purpose of the Study:

• To develop an Orthodontic Natural Language Processing (ONLP) model for structured information extraction from unstructured EDRs.
• To identify critical features influencing malocclusion using machine learning (ML) on extracted data.
• To enhance objectivity and consistency in orthodontic diagnosis and research.

Main Methods:

• Utilized a dataset of 7693 orthodontic patients for model training, testing, and validation.
• Developed an ONLP model using supervised (Named Entity Recognition) and unsupervised (K-means clustering) methods.
• Applied various ML models (Logistic Regression, Naive Bayes, Random Forest, XGBoost) to classify malocclusion and determine feature importance.

Main Results:

• The ONLP model achieved high accuracy (91%) in extracting orthodontic information.
• Supervised ML models showed 84% accuracy, excelling in Class I and III malocclusion identification.
• Key malocclusion features identified include crowding, overjet, arch perimeter discrepancy, spacing, midline deviation, and occlusal wear.

Conclusions:

• The developed ONLP model successfully automates orthodontic data extraction from EDRs.
• This approach enables advanced big data analytics for orthodontic research.
• The findings support data-driven improvements in orthodontic research and patient care.