Deep learning for age estimation from panoramic radiographs: A systematic review and meta-analysis

Area of Science:

• Radiology
• Artificial Intelligence
• Forensic Science

Background:

• Panoramic radiographs are standard for age estimation in clinical and forensic settings.
• Traditional methods rely on manual assessment of tooth development.
• Deep learning offers a potential automated approach for scalable age estimation.

Purpose of the Study:

• To systematically review and assess the performance of deep learning algorithms for age estimation using panoramic radiographs.

Main Methods:

• A systematic search of multiple databases (PubMed, Scopus, etc.) was conducted up to June 2024.
• Included studies used deep learning for age estimation and reported performance metrics.
• A meta-analysis was performed on 9 studies reporting mean absolute error.

Main Results:

• 42 studies were included, with 13 showing a low risk of bias.
• Accuracy metrics varied widely (27%-100%) for age bracket classification.
• The pooled mean absolute error for deep learning age estimation was 1.75 years.

Conclusions:

• Deep learning demonstrates potential as an adjunct tool for age estimation from panoramic radiographs.
• The mean absolute error of 1.75 years indicates promising performance.
• Methodological limitations require further investigation prior to clinical implementation.