Accurate age classification using manual method and deep convolutional neural network based on orthopantomogram...

Area of Science:

• Radiology
• Forensic Dentistry
• Artificial Intelligence

Background:

• Accurate age estimation is crucial for legal, clinical, and immigration contexts.
• Orthopantomograms (OPGs) are widely used for dental age assessment.
• Previous studies lack direct comparisons between manual and machine learning methods on large OPG datasets.

Purpose of the Study:

• To compare the performance of manual age estimation methods with end-to-end convolutional neural network (CNN) models.
• To evaluate both methods using a large dataset of 10,257 dental orthopantomograms.
• To determine if machine learning can surpass human accuracy in dental age classification.

Main Methods:

• Collected 10,257 OPGs for analysis.
• Developed logistic regression linear models for manual age estimation at legal thresholds (14, 16, 18 years).
• Trained an end-to-end CNN for direct dental age classification, analyzing permanent teeth or third molars.

Main Results:

• CNN models achieved higher accuracy than manual methods across all age thresholds.
• CNN accuracy: 95.9% (14 yrs), 95.4% (16 yrs), 92.3% (18 yrs).
• Manual method accuracy: 92.5% (14 yrs), 91.3% (16 yrs), 91.8% (18 yrs).

Conclusions:

• End-to-end CNN models demonstrate superior performance in dental age estimation compared to manual methods.
• AI-driven age classification using CNNs can surpass human capabilities.
• Machine learning may identify distinct features for age classification not apparent to human observers.