Machine learning approaches for sex estimation using cranial measurements.

Area of Science:

• Forensic Anthropology
• Medical Imaging
• Machine Learning

Background:

• Accurate sex estimation is crucial in forensic anthropology.
• Cranial measurements offer a potential basis for sex determination.
• Machine learning models can enhance classification accuracy.

Purpose of the Study:

• To apply support vector machines (SVM) and artificial neural networks (ANN) for sex estimation using cranial measurements.
• To compare the performance of SVM and ANN with logistic regression (LR).
• To evaluate the utility of attribute selection techniques in improving model accuracy.

Main Methods:

• Computed tomography (CT) images of 393 Bulgarian adults were analyzed.
• 3D coordinates of 47 landmarks were used to calculate 64 measurements and 22 indices.
• SVM, ANN, and LR models were trained and evaluated using 10x10-fold cross-validation.

Main Results:

• All methods achieved classification accuracies exceeding 95%.
• SVM demonstrated the highest accuracy (96.1% ± 0.5%), significantly outperforming ANN and LR.
• For interlandmark distances, attribute reduction via GeneticSearch improved SVM and ANN accuracy.

Conclusions:

• Machine learning models, particularly SVM, are highly effective for sex estimation from cranial data.
• SVM offers a robust and accurate method for sex determination in forensic contexts.
• The choice of dataset and attribute selection can influence model performance.