The Accuracy of Sex Identification Using CBCT Morphometric Measurements of the Mandible, with Different...

Area of Science:

• Forensic anthropology
• Radiology
• Computer science

Background:

• Accurate sex determination is vital in forensic identification.
• Conventional methods for sex estimation can be time-consuming.
• Advanced imaging and machine learning offer potential for faster, more precise identification.

Purpose of the Study:

• To evaluate the accuracy of volumetric and linear measurements from 3D cone beam computed tomography (CBCT) mandible images for sex identification.
• To compare the performance of different machine learning (ML) models in predicting sex from these mandibular measurements.

Main Methods:

• Collected CBCT scans from 104 males and 104 females.
• Generated 3D mandible images and extracted volumetric, surface area, and ten linear measurements.
• Applied five ML algorithms (including Gaussian Naive Bayes) for sex classification.
• Evaluated model performance using precision, recall, f1-score, and accuracy (p < 0.05).

Main Results:

• All measured mandibular parameters showed statistically significant differences between sexes (p < 0.05).
• The right coronoid-to-gonion linear distance exhibited the highest discriminative power.
• Gaussian Naive Bayes (GNB) achieved the best performance among the evaluated ML models.

Conclusions:

• 3D mandibular measurements from CBCT, analyzed with ML, provide a highly accurate method for sex determination.
• The GNB model demonstrated significant potential for reliable sex identification in forensic contexts.
• This approach offers a promising, accurate, and potentially faster alternative for forensic identification.