Deep Learning for Age Estimation and Sex Prediction Using Mandibular-Cropped Cephalometric Images: Comparative Model...

Area of Science:

• Forensic Odontology
• Artificial Intelligence
• Medical Imaging

Background:

• Mandibular structures are key for forensic identification using partial remains.
• Deep learning on cephalometric radiographs can predict age and sex for forensic and clinical use.

Purpose of the Study:

• Develop and evaluate a multitask deep learning framework for age and sex prediction.
• Analyze cropped mandibular regions from cephalometric radiographs.
• Compare deep learning models and preprocessing techniques for demographic prediction.

Main Methods:

• Utilized 340 Indonesian cephalometric radiographs (ages 8-40), cropping mandibular regions.
• Applied four preprocessing scenarios including Synthetic Minority Oversampling Technique and StandardScaler.
• Fine-tuned six pre-trained convolutional neural network backbones in a multitask framework.

Main Results:

• VGG16 model showed best age estimation (MAE 3.19 years) on original data.
• VGG16 achieved highest sex prediction accuracy (86%) with StandardScaler.
• VGG19 also demonstrated strong performance in sex prediction (82% accuracy).

Conclusions:

• Integrating mandibular cropping with deep learning improves demographic prediction from radiographs.
• AI-assisted forensic odontology can aid in disaster victim identification with partial remains.