Researchers based at the Children’s National Hospital in Washington have created a facial recognition tool using machine-learning technology to help clinicians diagnose rare genetic syndromes faster.

As described in The Lancet Digital Health, the tool was built based on a collection of 2800 photographs of children’s faces. Of these, 1400 had 128 known genetic conditions, such as Down syndrome and Noonan syndrome, and 1400 were controls without genetic syndromes matched for age, sex, and ethnicity.

Small differences in facial features are common in many rare genetic syndromes, but the differences can be subtle and difficult to pick up with the human eye alone. The machine learning program developed by the researchers is able to recognize small, common differences in facial features common to specific syndromes that a clinician may not pick up in a consultation, particularly if they do not have specialist knowledge.

The hope is that this software will help clinicians come to a diagnosis more quickly, a process that can be very time consuming and stressful for the affected children and their families.

“We built a software device to increase access to care and a machine learning technology to identify the disease patterns not immediately obvious to the human eye or intuition, and to help physicians non-specialized in genetics,” said Marius George Linguraru, professor and principal investigator at the Sheikh Zayed Institute for Pediatric Surgical Innovation at Children’s National Hospital, as well as lead author of the paper.

“This technological innovation can help children without access to specialized clinics, which are unavailable in most of the world. Ultimately, it can help reduce health inequality in under-resourced societies.”

The children photographed for the dataset come from many countries around the world and were 47% female and 53% male. Most were White (56%), but other ethnicities were also represented with 15% African, 15% Hispanic and 13% Asian children making up the remainder of the group.  The children included ranged from newborn up to 20 years of age.

The overall accuracy of the model for correctly detecting a genetic syndrome was 88%, with a 90% sensitivity and 86% specificity. The tool was designed to account for normal facial variation in a given population, but accuracy of prediction was greater in White and Hispanic children at 90% and 91%, respectively. Accuracy in African and Asian groups was lower at a respective 84% and 82%.

“Facial appearance is influenced by the race and ethnicity of the patients. The large variety of conditions and the diversity of populations are impacting the early identification of these conditions due to the lack of data that can serve as a point of reference,” said Linguraru. “Racial and ethnic disparities still exist in genetic syndrome survival even in some of the most common and best-studied conditions.”

Sex and age did not seem to significantly impact the accuracy of the model, although accuracy was slightly lower for children younger than two years (86%) versus those aged 2–5 years (89%).

The tool has been designed using machine learning technology so has the advantage that it can improve or ‘learn’ as more data is added. The authors hope that by collecting more data it should become more accurate over time.

It was announced in July that MGeneRx, a spinoff from BreakThrough BioAssets LLC, will licence the software from Children’s National Hospital.

“The social impact of this technology cannot be underestimated,” said Nasser Hassan, acting CEO of MGeneRx. “We are excited about this licensing agreement with Children’s National Hospital and the opportunity to enhance this technology and expand its application to populations where precision medicine and the earliest possible interventions are sorely needed in order to save and improve children’s lives.”