Facial analysis tool detects 128 genetic diseases in children
Researchers at Children’s National Hospital in Washington, DC, have developed a new biometric analysis tool that can screen for 128 genetic diseases in children simply by examining their facial features.
Although the tool has not yet been approved for patient care, it has been cleared by MGeneRx to seek regulatory and marketing approval for its use as a diagnostic and screening tool.
The machine learning model stratifies patients based on their risk for genetic disease, said Marius Linguraru, senior researcher at the hospital’s Sheikh Zayed Institute for Pediatric Surgical Innovation. It was developed by looking at photos of patients of all races, ages and genders with genetic diseases to see how these conditions manifested on the face.
It is intended for use as a population screening tool that can point to a patient who might benefit from further referral to a genetic counselor, he said.
Not a Modern Healthcare subscriber? Sign up today for $ 1.
The tool, which has been in development since 2012, has been formed with both images of children without genetic syndrome to learn which variations in facial features are normal and expected, as well as images of patients with genetic disorders diagnosed for identify key markers, he said. noted.
He was formed with 2,800 children from 28 countries with 128 different genetic diseases, including Williams-Beuren syndrome, Cornelia de Lange syndrome and Down syndrome, he said, from newborns to babies. patients in late adolescence. The control group consisted of 1,400 images. The photos were retrospectively retrieved from three publicly accessible databases, as well as the archives of Children’s National and photos taken on a smartphone at the hospital.
The deep learning architecture is structured into three neural networks: network A, which standardizes photos; Network B, which detects facial morphology; and Network C, which estimates the risk of genetic syndrome.
The team grouped the photos into smaller subsections for cross-validation and recycled the tool, which Linguraru said showed improvements. These improvements were not surprising, because as the machine learning algorithm sees more patients, it “becomes better at determining” relevant facial features.
âThe more data he has, the smarter he gets and the more focused he is,â he said.
According to a study by researchers published in The Lancet Digital Health earlier this month.
The machine learning model performs “an in-depth analysis of facial patterns,” Linguraru said. The algorithm examines âeverything on the faceâ, such as the line and indentation patterns on the face, the distance between the eyes and the length of the philtrum. It is “looking for changes” in these measures, he said.
It then measures those distances and analyzes the patterns of those lines, returning a result in about a second that exposes the patient’s risk of genetic disease. While it can be used on people of all ages, Linguraru said development is focused on pediatric patients under the age of 21.
Examining a global population that includes people of all races and mixed race people, Linguraru said the team has made an effort to sidestep potential biases. syndromes in non-Europeans.
The team tried to have “so much data [that’s] as diverse as possible “to ensure that the tool would be able to detect important facial features of all patients, he said. There are also differences between patterns and measurements in different ethnic groups that the tool had to be trained to be aware, he said.
However, the researchers noted in the Lancet study that the tool was more accurate in white and Hispanic populations than African and Asian populations – 90 percent and 91 percent accuracy in white and Hispanic patients, compared with 84 percent and 82 percent accuracy in populations African and Asian, respectively.
They also said they used a broad definition of racial or ethnic categories that may not distinguish between the finer variations within populations – for example, the Asian group included patients from China, Japan, d ‘India and Thailand. âThis large group of individuals could be another reason for the difference in performance,â they wrote.
The accuracy was similar in male and female children, and slightly better in children aged 2 to 5, according to the study.
The next step in the tool is clinical validation using a prospective patient cohort and determining an âadequate functioning thresholdâ in the classifier that provides âclinically significantâ sensitivity and retains acceptable specificity. , wrote the researchers.
Linguraru has taken care to note that the tool is not intended for use by a geneticist. Instead, it is for “that first level of care”, like a primary care physician, who can refer the patient further.
Providing earlier detection of genetic disorders is beneficial because it dramatically improves survival rates as patients can see genetic specialists earlier and receive preventive care, Linguraru said.
Other researchers have used biometric analysis for medical diagnosis, including a German team that published an article in 2003 to determine whether a computer can recognize disease-specific facial patterns, a team from the University of Oxford and researchers in China who have developed a facial feature recognition tool. of Turner syndrome.
Additionally, the Boston-based FDNA has marketed facial recognition software for use in genetics, Face2Gene, which in a 2015 study published in Clinical genetics had an accuracy of 87 percent.
However, a more recent study from the start of the year in the American Journal of Medical Genetics found an overall diagnostic yield of 57 percent, which rose to 82 percent when cases diagnosed with syndromes not recognized by Face2Gene were removed.
Marketing is on the move
All intellectual property relating to the Children’s National Hospital tool has been licensed to MGeneRx, which will make it into a commercial product.
Nasser Hassan, acting CEO of MGeneRx, said the company plans to use the technology in a number of ways, including as a smartphone app, as it would be “the most direct route to a pediatrician.” and other health professionals. He also plans to work with different hospitals to integrate the tool into their electronic medical records, allowing pediatricians to use it directly in their offices for screening along with other medical information, he added.
The main focus right now is to use it as a screening tool, and that’s what he’s working on, in addition to working with the United States Food and Drug Administration to further develop the tool. MGeneRx also sees the potential of the technology as a rare disease diagnostic tool – something it is actively exploring, Hassan said.
MGeneRx is currently raising funds to further develop the tool and move towards commercialization and is seeking government, private and philanthropic funding, he said. The company is also refining the tool so that it is fully functional on all different mobile devices, he added.
It would also likely be cleared by a hospital or healthcare system for use, rather than a direct-to-consumer application – although it could potentially be used directly. Hassan noted that for such a sensitive subject, patients would want to go through doctors for proper advice.
The company is working to market it first in the United States, hopefully within a year, and then expand it to other countries, including low- and middle-income countries where children cannot. not have access to genetic counseling, he said. It would be available wherever there is internet access and adapted to the environment, he continued.
âWe are really developing it because of the social impact of a product like this,â Hassan said.