Biological Gender Estimation from Panoramic Dental X-ray Images Based on Multiple Feature Fusion Model

Abstract

In a catastrophe, the conventional biological characteristics of the victims will be destroyed. Forensic odontology is the main method to identify the victims. Estimating the gender of the victims has a significant meaning and can greatly help identify the victims. In this paper, we propose a new automatic method to the gender estimation from panoramic dental X-ray images based on improved convolutional neural network with multiple feature fusion module. Our dataset includes 19,976 panoramic dental X-ray images from Chinese patients. The method we propose can estimate 142 images per second on the conventional computing equipment and it achieves state-of-the-art performance, accuracy of 94.6% ± 0.58%, in our dataset. Our model is interpreted by perturbation-based forward propagation approaches, and the results show that focus of our method on the area of mandible and teeth is reliable which is in accordance with forensic practice.