Automatic frontal sinus recognition in computed tomography images for person identification

Area of Science:

• Forensic Anthropology
• Medical Imaging
• Biometrics

Background:

• Biometric person identification often relies on hard tissues like teeth and bones when other methods are not feasible.
• Computed tomography (CT) imaging provides detailed anatomical data for identification purposes.

Purpose of the Study:

• To develop and evaluate a novel method for person identification using frontal sinus features extracted from CT images.
• To automate the segmentation of the frontal sinus in CT scans for feature extraction.

Main Methods:

• Automatic segmentation of the frontal sinus from CT images using a newly developed algorithm.
• Extraction of shape features from the segmented frontal sinus using the Beam Angle Statistics (BAS) method.
• Person identification by comparing extracted features using L2 distance.

Main Results:

• The proposed frontal sinus recognition method achieved an identification accuracy of 77.25% on a dataset of 310 CT images from 31 individuals.
• Automatic frontal sinus segmentation demonstrated high accuracy, with a mean Cohen Kappa coefficient of 0.8852 compared to manual segmentation.

Conclusions:

• Frontal sinus features extracted from CT images represent a viable biometric modality for person identification.
• The automated segmentation and feature extraction method shows promise for forensic and identification applications.