Sex Identification Using Computed Tomography Images of the Human Skull

Reconstructing human faces from skull images is a vital task in fields such as forensic science, archaeology, and medicine, contributing to the identification of unknown individuals and the creation of detailed biological profiles, including sex, age, and ancestry. Traditional methods for biological profiling are labor-intensive, reliant on specialized expertise, and time-consuming. This study explores the integration of machine learning to address these challenges, focusing on the development of a robust model for sex identification using convolutional neural networks (CNNs). A transfer learning approach was employed, utilizing pre-trained CNNs on the ImageNet database to extract features and training a task-specific classification layer on a dataset of computed tomography (CT) skull images. Three prominent architectures-Oxford VGG, Google Inception, and Microsoft ResNet-were evaluated. The ResNet152 model achieved the highest performance, with an accuracy of 93.46% on the test dataset, matching state-of-the-art results. This work highlights the potential of machine learning to improve efficiency, accuracy, and scalability in biological profiling applications.Clinical relevance- The integration of machine learning in sex identification enhances the accuracy and efficiency of biological profiling, enabling faster identification of individuals in forensic and medical contexts. By reducing reliance on highly specialized professionals and time-intensive methods, this approach supports resource-constrained environments and time-sensitive scenarios, such as disaster victim identification or criminal investigations. Additionally, its scalability and precision can enhance the integration of technological advancements into practical clinical applications.