Age determination from images in adolescents

Age Determination From Images In Adolescents Age determination in forensic medicine is requested in various cases of civil and criminal law. Although there are many methods for age determination, traditional methods are not sufficient for determining age based on images. Tanner staging can be used for age determination in cases that consist of photographs and video recordings, but this method has limitations. In recent years, the photo-anthropometry method has been used to determine age by indices calculated from facial measurements for age determination. Our study utilized this method to determine whether individuals in the image were under or over 18 years of age. The study included frontal and lateral photographs of 637 volunteers, 330 females, and 307 males aged between 15 and 20 years old. Within the scope of the study, 9 frontal and 4 lateral measurements were taken by marking 17 anatomical points in frontal photographs and 4 anatomical points in lateral photographs. Based on the facial measurements, 22 indices were analyzed to determine their correlation with the age and gender of the participants, then machine learning models were developed to group individuals by age and gender using calculated indices. Statistical analyses were performed using IBM SPSS Orange Data Mining and R Studio. Inter-observer reliability was moderate or above in all measurements, and the analysis proceeded with the calculated indices. There were significant differences in age groups between males and females in 10 indices in the frontal photograph and 3 in the lateral photograph. The machine learning models were developed and used to group individuals by age and gender using all indices. The SVM model correctly classified the gender of 80.8%, and the age of 79.9% of the individuals. As a result, the models we developed can accurately classify individuals based on age and gender using calculated indices. Our study evaluated the accuracy of age classification by indices derived from standardized photographs, furthermore, it needs to be tested with similar studies on random photographs.