Brazilian researchers have developed a machine learning system that can determine an individual's gender based on panoramic radiographs (wide-field dental X-ray images of the entire mouth). The system was 96% accurate when the images had good resolution and the individual was over 16 years old. For younger individuals, the accuracy was lower. The results of their study: Journal of Forensic Science.
When forensic experts need to identify an individual based on a human remains, one of their first goals is to determine the gender of the deceased. Throughout history, various methods have been used to do this, many of which are based on the analysis of bones, their structure and relationships. Forensic methods that extract the necessary information from bones are particularly important, as bones are often the best-preserved part of the body.
In some cases, forensic experts only have to deal with the jaw and teeth. Forensic odontology techniques are of paramount importance, especially in mass accident situations and situations where the various stages of human decomposition, such as charred, macerated, or skeletal, are to be analyzed.
As in many other areas of human activity, artificial intelligence (AI) is finding its way into this field. AI-based methods can help dentists interpret images more effectively, reduce human error, and make faster decisions. AI-based techniques can also potentially capture information that is not easily detectable by the naked eye or traditional methods.
Study author Ana Claudia Martins-Chiconer and her colleagues wanted to explore whether they could develop an AI-powered tool that could accurately estimate an individual's gender based on a panoramic radiograph — a wide-field dental X-ray that captures the entire mouth, including the teeth, jaw, and surrounding structures, in one comprehensive picture.
The researchers collected 207,946 panoramic radiographs and corresponding reports from 15 clinical centers in São Paulo, Brazil. The panoramic radiographs were obtained using four different devices. Fifty-eight percent of the patients were female. All patients were alive at the time the radiographs were taken. Forty-three percent of the patients were missing up to four teeth, and 5 percent were missing 16 or more teeth.
The study authors organized each patient's data into a database and trained two machine learning algorithms to estimate gender based on panoramic radiographs. One algorithm was a convolutional neural network and the other was a residual network. A convolutional neural network is a type of deep learning model that learns a hierarchy of features in a set of input images. A residual network is a type of deep learning model that uses shortcuts to pass information between layers more efficiently.
Results show that after optimization, both types of algorithms showed similar accuracy in estimating gender. The main factor affecting accuracy is the photo resolution – the higher the resolution, the higher the accuracy. The second most important factor is age. For patients between 20 and 50 years old, the system's accuracy was above 97%. For patients over 70 years old, the accuracy was just below 95%.
For patients aged 6 to 16 years, the system estimated gender with 87% accuracy, but only 74% accuracy for children under 6. For individuals aged 16 and over, the overall accuracy was 96%.
“This study demonstrates the effectiveness of an AI-driven tool for sex determination using PR. [panoramic radiographs] “Our findings highlight the role that image resolution, age, and gender play in determining algorithm performance,” the study authors conclude.
The study presents a new software tool that can determine an individual's gender based on panoramic X-rays. However, it is important to note that all images used in the study were from living people. Images from human remains that are already in an advanced stage of decomposition may have different validity.
The paper, “Deep Learning for Gender Determination: Analysis of Over 200,000 Panoramic Radiographs,” was written by Ana Claudia Martins Ciconelle, Renan Lucio Berbel da Silva, Jun Ho Kim, Bruno Aragão Rocha, Dênis Gonçalves dos Santos, Luis Gustavo Rocha Vianna, Luma Gallacio Gomes Ferreira, Vinícius Henrique Pereira dos Santos, Jefferson Orofino Costa and Renato Vicente.
