With or without human interference for precise age estimation based on machine learning?

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Forensic Odontology

Background:

• Dental age estimation is crucial for forensic medicine, orthodontics, and pediatrics.
• Current methods are subjective, time-consuming, and require specialized expertise.
• Machine learning offers a potential solution to enhance objectivity and efficiency.

Purpose of the Study:

• To compare two convolutional neural network (CNN) approaches for dental age estimation.
• To evaluate the impact of feature extraction methods (manual vs. automated) on model performance.
• To determine the efficacy of fully automated deep learning for dental age estimation.

Main Methods:

• Development of an automated dental stage evaluation (ADSE) model using manually defined features.
• Development of an automated end-to-end dental age estimation (ADAE) model with autonomous feature extraction.
• Comparison of ADSE, ADAE, and manual dental age estimation (MDAE) model performance using Mean Absolute Error (MAE).

Main Results:

• The ADSE model showed an MAE of 0.17 stages for classification but an unsatisfactory MAE of 1.63 years for age estimation.
• The ADAE model achieved a significantly lower MAE of 0.83 years, approximately half that of the MDAE model.
• Fully automated feature extraction demonstrated superior performance compared to manual or semi-automated methods.

Conclusions:

• Deep learning models with autonomous feature extraction excel in dental age estimation.
• Automated methods enhance accuracy and objectivity in dental age assessment.
• Machine learning, particularly deep learning without human interference, shows significant promise for medical imaging applications.