Age and Sex Prediction from Cervical Vertebrae Cephalogram Image Using Convolutional Neural Network Model

Understanding the growth process is crucial in determining the optimal timing for dentofacial orthopedic treatment, to correct bone disharmony. To reduce radiation exposure, a relatively new method of assessing bone maturity is using the neck bone (cervical vertebrae). In the forensic field, age is one of the important factors in determining a person's identity. Estimates of age and sex are used in human identification in forensics and criminal and civil proceedings. In cases where an intact skull cannot be found, the analysis of cervical vertebrae alone can help in determining sex and age. However, the manual analysis process is time-consuming and requires extensive measurements. Therefore, automated analysis is used. This study experimented with the usage of cephalogram and cervical vertebrae images to build a sex and age prediction model using convolutional neural network (CNN). This study aims to identify performance differences between the usage of said images. Standardized hyperparameters were used across experimentation to ensure fair comparisons. Our study revealed that models built using cervical vertebrae images output competing performance compared to those built using a full cephalogram image. The best-performing models achieved 94% accuracy in sex prediction and a mean percentage error (MAPE) of 16,36% in age prediction.