Sex estimation from morphometry of lumbar vertebrae (L4-L5): A machine learning approach

Sex estimation is the first and most important step in establishing a reliable biological identity during the examination of skeletal remains. For sex estimation, the pelvis and cranial bones are primarily preferred. In cases where the pelvis bones are not available, vertebrae are also utilised. The present study focuses on the effectiveness of the fourth-fifth lumbar vertebrae in sex identification within the Turkish population using machine learning.

This study was conducted by screening retrospective computed tomography images of individuals admitted to Karaman Training and Research Hospital. DICOM images of the fourth-fifth lumbar vertebrae of 540 individuals aged between 18 and 79 were used. Measurements included Vertebral Length, Vertebral Body Anterior Height, Transverse Process Length, Spinous Process Length, Vertebral Foramen Anterior-Posterior Diameter, Vertebral Foramen Transverse Diameter, Vertebral Body Anterior-Posterior Diameter, and Vertebral Body Transverse Diameter. Eight different machine learning algorithms were employed for sex estimation.

Among the machine learning algorithms, the logistic regression algorithm yielded the highest accuracy, reaching 91.3% for the fourth lumbar vertebra and 90.7% for the fifth lumbar vertebra. Moreover, the most effective parameters for sex estimation in both vertebrae were found to be Vertebral Body Anterior-Posterior Diameter and Vertebral Body Anterior Height.

The findings of this study demonstrate that the fourth-fifth lumbar vertebrae exhibit dimorphism and can be used for sex estimation. We consider that this study may be beneficial to forensic researchers and anthropologists in terms of sex estimation and to clinicians and regional surgeons in terms of morphometry.