Age group prediction with panoramic radiomorphometric parameters using machine learning algorithms.

Area of Science:

• Dentistry
• Radiology
• Machine Learning

Background:

• Accurate age estimation is crucial in forensic odontology and clinical practice.
• Radiomorphometric parameters from panoramic radiographs offer a non-invasive method for age assessment.
• Machine learning (ML) algorithms can potentially enhance the accuracy and efficiency of age estimation.

Purpose of the Study:

• To investigate the relationship between 18 radiomorphometric parameters on panoramic radiographs and age.
• To develop and evaluate five ML algorithms for estimating age groups in individuals with permanent dentition.
• To identify key radiomorphometric parameters predictive of different age groups.

Main Methods:

• Analysis of 471 digital panoramic radiographs from Korean individuals (209 men, 262 women).
• Application of five ML models: linear discriminant analysis, logistic regression, kernelized support vector machines, multilayer perceptron, and extreme gradient boosting.
• Classification into three (20-year gap) and six (10-year gap) age groups, with Fisher discriminant analysis for visualization.

Main Results:

• High Area Under the Curve (AUC) values were achieved for young (10-19 years) and older (50-69 years) age groups across ML models (AUCs 0.85-0.90).
• Adult age groups (20-49 years) showed moderate AUC values (approx. 0.73).
• Feature analysis identified L-Pulp Area for younger individuals and L-Crown, U-Crown, L-Implant, U-Implant, and Periodontitis for older individuals.

Conclusions:

• Acceptable linear and nonlinear ML models were established for dental age estimation using radiomorphometric parameters.
• Automated ML models can effectively distinguish young and elderly individuals due to age-related radiomorphological characteristics.
• This approach provides a non-invasive, comprehensive, and accurate method for age group estimation.