Children’s National Hospital’s AI enables rapid genetic screening.

Artificial intelligence (AI) machine learning is increasingly being used as a diagnostic tool for health care, biotechnology research, medical care, and life sciences. A new study published today in The Lancet Digital Health by researchers at the Children’s National Hospital in Washington, DC unveils an AI deep learning tool that can detect the risk of genetic syndromes in children with 88 percent accuracy.

“Genetic syndromes can be associated with severe cardiovascular, immune, endocrine, and neurodevelopmental risks, and thus have an impact on the quality of life of patients and their families,” wrote the researchers affiliated with the Sheikh Zayed Institute for Pediatric Surgical Innovation at Children’s National Hospital in Washington, DC, and George Washington University.

The researchers designed the AI deep learning architecture to consist of three artificial neural networks to perform image standardization, facial morphology detection, and genetic syndrome risk evaluation.

“Delays in the diagnosis of genetic syndromes are common, particularly in low and middle-income countries with limited access to genetic screening services,” the scientists wrote. “We, therefore, aimed to develop and evaluate a machine learning-based screening technology using facial photographs to evaluate a child’s risk of presenting with a genetic syndrome for use at the point of care.”

The data used for the study include 128 different genetic conditions such as Down syndrome, Williams-Beuren syndrome, Cornelia de Lange syndrome, 22q11.2 deletion, and Noonan syndrome.

“In this retrospective study, we developed a facial deep phenotyping technology based on deep neural networks and facial statistical shape models to screen children for genetic syndromes,” wrote the researchers.

To train the deep neural networks, the researchers used a dataset of 2,800 retrospective facial photographs of children that included 1,400 children that were diagnosed with 128 genetic conditions, as well as 1035 syndromic photographs from the Children’s National Hospital and 365 from other datasets from other research studies, Face2Gene, and the Atlas of Human Malformations in Diverse Populations of the National Human Genome Research Institute.

“Facial appearance is key in a geneticist’s evaluation of people with suspected genetic syndromes,” explained the researchers. “However, primary care physicians, who are not trained to identify dysmorphology in diverse populations, often miss subtle indicators of genetic conditions. Additionally, cautious clinicians might refer healthy children with an atypical facial appearance to expensive and unnecessary genetic evaluations.”

The researchers reported that their AI deep learning solution could detect genetic syndrome with a high degree of accuracy of 88 percent for the general population.

“This provides a substantial improvement over the reported accuracy of trained pediatricians to identify well studied conditions such as Down syndrome (64% accuracy from physical examination),” reported the researchers. “Our results demonstrate the feasibility of our method, and the potential to improve the early detection of genetic syndromes.”

With this proof-of-concept, the researchers identify clinical validation using a prospective patient cohort as the next step. According to the scientists, their AI technology is designed to be deployed at the point of care such as primary care facilities, pediatric health care clinics, and maternity wards, via a smartphone app to enable early detection of genetic syndromes. It is not intended to replace genetic testing, but rather as an assistive tool for clinicians for preventive care, diagnosis, and screening.

“This genetic screening technology could support early risk stratification at the point of care in global populations, which has the potential to accelerate diagnosis and reduce mortality and morbidity through preventive care,” the researchers concluded.

Copyright © 2021 Cami Rosso All rights reserved.

Share This