In 2015, Dr. Antonio Porras capped a dozen years of higher education with a PhD in medical computing from Pompeu Fabra University in Barcelona, Spain. After graduation, his training didn’t produce the satisfaction he expected.
“I was creating all these cool [computing] methods and publishing them, but nobody was using them because I was outside a hospital,” Porras recalled. “I didn’t have any medical exposure.”
Porras wanted to work in a clinical environment, where he could help patients more directly. In 2019, he followed his desires, entering post-graduate training at the Foundation for Advanced Education in the Sciences, which is part of the National Institutes of Health in Bethesda, Maryland. A critical career turn came at Children’s National Hospital, in Washington D.C., where he studied management of pediatric neurosurgery patients.
The work opened a new world for Porras. He had earned his PhD studying cardiovascular imaging, which focused almost entirely on older adults. In his new role, he was interested in using artificial intelligence (AI) to detect disease in pediatric patients, but he quickly recognized that imaging and AI research with kids is a lot more difficult than with adults, for one simple reason.
“The big difference comes from the fact that kids are growing and changing very fast,” Porras said. “If you look at a child today who is three weeks old, they will look very different than when they are three months old, even though there is a very little time span in between.”
Using artificial intelligence to ferret out craniofacial disease clues in kids
That insight triggered for Porras an ongoing interest in finding ways to use AI and other imaging technologies to understand childhood development. Initially, he worked on studying craniofacial anomalies associated with genetic syndromes in children.
“Babies often experience nearly imperceptible physical anomalies that are signs of a genetic disease but that changes quickly over different points in time,” Porras explained. “These anomalies depend on factors such as age or race. The same anomaly will not look the same for two children with different age, sex and race.”
In contrast, Porras said, “With adults, any major change you see in a relatively short time is often caused by a pathology that is making the body change,” Porras said. “With kids, major changes are expected, and they are normal.”
Now an assistant professor in the Department of Biostatistics and Informatics at Colorado School of Public Health, and director of research in the Department of Plastic and Reconstructive Surgery at Children’s Hospital Colorado, Porras is a leader in efforts to build reservoirs of data about pediatric craniofacial anomalies like craniosynostosis, a condition caused by a baby’s skull bones fusing too early. The goal: help surgeons hone and personalize their plans for the delicate corrective procedures that are required and deepen understanding of what causes these diseases.
Porras has also conducted and published important research in the field. The subjects of his co-authored work include machine learning to screen for genetic syndromes in children; advanced imaging to detect bone density changes that are flags for increased pressure on the brain; and improved 3D photogrammetry that rapidly assesses craniosynostosis.
A wealth of research in a field ripe for investigation
Ultimately, Porras’s current work rests on the professional path he took after leaving Barcelona and learning about the clinical and translation realm. Early in his post-graduate studies, he worked with Children’s National Hospital and the National Human Genome Research Institute (NHGRI) on a project to understand how age and other factors like race and sex can influence craniofacial anomalies – those of the bones of the skull and face – and how they are associated with various genetic syndromes.
“We were trying to see how [a child’s] development was changing [their facial characteristics],” Porras said. It was a very important question, he said, because the number of experts who can recognize craniofacial anomalies associated with genetic syndromes around the world is insufficient, especially in developing and rural regions.
He noted, for example, that one in three babies born with a genetic syndrome are not identified at birth, simply because the early signs of many disorders are very subtle, and only a small percentage of the world’s population has access to prenatal screening for a few genetic syndromes.
To address the problem, Porras’s collaborators at NHGRI built a database of pictures of babies with and without genetic syndromes in the Atlas for Human Malformation Syndromes in Diverse Populations.
“Their idea was to create a data set that would not be static, as with adults, but would be a developmental reference that could account for age and race,” he said. Based on that data set, Porras and his team designed machine learning methods that could identify genetic syndromes with 88% accuracy in the general pediatric population, he added. That success rate was achieved from a list of 2,800 patients with 128 different syndromes from 28 countries .
“Our results demonstrate the feasibility of our method, and the potential to improve the early detection of genetic syndromes,” Porras and his co-authors wrote in a summation of the study published in Lancet Digital Health in 2021.
Probing the roots of disease
Porras is also keenly interested in probing the genetic roots of craniofacial anomalies. For example, he’s helped to build a biobank – a repository of data gleaned from patients treated by the Craniofacial Program at Children’s Colorado, which has one of the highest volume of cases in the United States. He is joined in that effort by Dr. Brooke French, MD, co-director of the Children’s Craniofacial Clinic and principal investigator of this initiative, and Cristan Carter, senior research professional with the Department of Surgery at the University of Colorado School of Medicine.
Data from the biobank will drive collaboration with basic scientists to investigate genetic variations in groups of patients with specific craniofacial anomalies, Porras said. For example, the Craniofacial Program is collaborating on such a study with Dr. Tamim Shaikh, MD, PhD, director of the Human Medical Genetics and Genomics Graduate Program at the University of Colorado Anschutz Medical Campus.
Shaikh will do the genetic analysis, bolstered by photographic and CT imaging of kids with and without the anomalies, Porras explained. The idea is to not only discover the genetic variations that give rise to disease, but also link that information to the physical changes that result. With that information, surgeons will be able to optimize the treatment they provide patients rather than adhere to more general approaches.
With expanded data sets about patients, “we have information at every single stage of the process so we can start building these connections between the cause of the disorder and its best treatment, based on quantitative knowledge,” Porras said.
Looking to the future, Porras and his team are also considering how to create “synthetic data” from the image repository of patients with craniosynostosis. That would allow researchers at other institutions to conduct further investigations while also protecting patient privacy.
“Democratizing craniofacial research is probably the next step in the field,” Porras said. “It is letting others use data like ours to answer other questions. Right now we cannot do that, but we hope we can do it in the next couple of years.”