Oguzhan Ozturk; Oguzhan Harmandaoglu; Seren Kaya; Yusuf Secgin; Deniz Senol; Serdar Colakoglu & Omer Onbas

SUMMARY: The skull is one of the most dimorphic and anatomically informative bones for sex estimation and shows resistance to taphonomic processes. This study aims to estimate sex using machine learning (ML) algorithms based on morphometric measurements of the optic canal (OC)—a clinically significant canal within the sphenoid bone that transmits the optic nerve and ophthalmic artery. This retrospective study was conducted on CT from 260 adults (130 females and 130 males, aged 18–65). The images were obtained from the PACS archive of the Department of Radiology, Faculty of Medicine, Düzce University, covering the years 2019 to 2025. Sixteen bilateral morphometric parameters of the temporal bone were measured in axial and coronal planes. Data were analysed using various ML algorithms, and classification performance was compared. On the 20 % test set, ML models achieved over 81 % accuracy; Logistic Regression performed best with 90 %. In 10-fold cross-validation, all algorithms exceeded 74 %, with LR again reaching the highest at 89 %. Decision Tree yielded the lowest accuracy. SHapley Additive exPlanations (SHAP), which facilitates interpretable machine learning, revealed that the right-sided OC–midsagittal distance had the greatest predictive impact. Morphometric data from the OC provide high accuracy and strong potential for sex estimation. The study also highlights sex- and population-based variation in OC position. These findings may be relevant in clinical and forensic contexts, particularly in forensic anthropology, ophthalmology, and legal medicine.

KEY WORDS: Sex estimation; Machine learning algorithms; Sphenoid bone; Optic canal.

OZTURK, O.; HARMANDAOGLU, O.; KAYA, S.; SECGIN, Y.; SENOL, D.; COLAKOGLU, S. & ONBAS, O. Sex estimation based on radiological measurements of the canalis opticus and surrounding structures using machine learning algorithms. Int. J. Morphol., 43(6):2155-2162, 2025.