Automated Gender Determination Using Dental Features and Convolutional Neural Networks

Gender determination is a critical task in forensic odontology, aiding in using dental features extracted from digital panoramic radiographs and Convolutional Neural Networks (CNNs). The study addresses the challenge of manual analysis by leveraging machine learning to achieve higher accuracy and efficiency. A comprehensive dataset comprising annotated radiographic images was utilized, ensuring representation across genders and age groups. The images underwent rigorous preprocessing steps, including normalization, resizing, and data augmentation, to enhance model generalization and reduce overfitting. The CNN architecture, based on transfer learning, was fine-tuned on the dataset, replacing the output layer for binary classification. Training was optimized using a combination of cross-entropy loss and the Adam optimizer, with early stopping to prevent overfitting. The proposed model achieved significant performance metrics, including an accuracy of over 90%, along with high sensitivity, specificity, and F1-score. Grad-CAM visualizations were employed to interpret the network’s decision-making process, highlighting the critical dental regions contributing to gender classification. This interpretability ensures transparency and facilitates acceptance in clinical and forensic domains. The study demonstrates that the proposed automated approach not only reduces the subjectivity inherent in manual methods but also achieves rapid, reproducible, and scalable results. The findings underline the potential of AI-driven solutions in forensic science and dentistry, offering a promising tool for gender determination and related applications. This study will focus on expanding datasets and refining the model to address variations in image quality and population diversity in future.