New algorithm for the forensic determination of age from chest radiographs using artificial intelligence

Abstract

Age determination by forensic investigators can be enhanced through artificial intelligence, which has demonstrated significant potential in the analysis of medical images. The study aims to provide an accurate and reliable method for age estimation from chest radiographs and to test the deep neural network’s performance for age classification. This retrospective study included chest x-rays of 562 Egyptian patients obtained from the Department of Diagnostic and Interventional Radiology, Faculty of Medicine, Cairo University. They were analyzed according to age. We tested the deep convolutional neural network for age determination using a new algorithm. Mean absolute error of age prediction was 7.4 years, final loss in age prediction was 6.9, correlation coefficient between actual age and predicted age was 0.844 with P value < 0.001 which is considered statistically significant. Artificial intelligence helps in age determination with some limitations that need further studies. We recommend establishing a national dataset of radiographs and using artificial intelligence in medico-legal investigations requiring positive identification.