How is machine learning helping us more accurately identify intracranial pressure to better plan surgical interventions in kids with craniosynostosis?
Deciding when to operate on children with craniosynostosis is a delicate dance that doesn’t hinge on one factor alone. Surgeons weigh the risks of anesthesia, blood loss and reoperation rates associated with surgery against intracranial pressure, brain development and age. Still, some of it is educated guesswork, since symptoms are not consistently present. That leaves just one way to really know whether a baby has enough intracranial pressure to cause long-term problems. Doctors must drill into the skull and measure — a procedure that is usually avoided because of its risks.
For years, Antonio R. Porras, PhD, Research Director of Pediatric Plastic and Reconstructive Surgery at Children's Hospital Colorado, has combined his medical knowledge with computational and data science expertise to reduce guesswork in the process. His research uses data and artificial intelligence (AI) to guide physicians toward safer and more precise surgical timelines.
The standard of care for craniosynostosis
Children are born with sutures in the skull that allow for cranial expansion as the brain grows. In some children, one or more of these sutures close prematurely, constraining the brain as it increases in size — a condition called craniosynostosis. This causes intracranial pressure that can have unpredictable effects, including vision loss, hearing impairment and developmental delays. The condition is usually treated through surgical interventions.
“What happens during an open surgery is that surgeons literally break the cranial bones and reshape them to create space for the brain and correct the abnormal shape,” Dr. Porras says. “It can be pretty intense, so that's why they try to wait until they are 5 or 6 months old instead of taking a baby that may be diagnosed at the age of 2 weeks into this surgery.”
Some infants show signs of pressure that indicate surgery may be needed sooner, including changes in eating, cognitive or developmental delays or hydrocephalus, a buildup of fluid in the brain. But because these young kids are developing fast and can quickly adapt to pressure changes, signs of increased intracranial pressure are often subtle and easy to miss, unlike in the adult population.
While these symptoms provide an idea that pressure may be rising to a dangerous level, that’s not always the case. In fact, doctors typically don’t know how high the pressure is until they either measure it or find other signs inside the patient’s skull during surgery.
That leaves them with the difficult task of deciding when the risks associated with intracranial pressure outweigh those of early surgery.
Creating a more accurate measurement
Dr. Porras’ latest work aims to offer doctors a clearer perspective on the presence of intracranial pressure, based on bone density and thickness — two measures that can be taken through a simple CT scan.
“I noticed when I was going through these surgeries that when the brain has pressure, the bone has very subtle changes that you can only see when you operate on the kid. You can actually see that the density of the bone decreases, but this is something that is very, very hard to see because of the subtlety of the changes and their dependence on the specific age of the patients,” Dr. Porras explains.
To understand the connection between these variables and intracranial pressure, Dr. Porras and his team conducted a retrospective study to quantify cranial bone thickness and mineral density from CT images of children with chronic intracranial pressure. They compared these to data from children without the condition and found that in kids with chronic intracranial pressure, bone density was lower, and bones were typically thicker.
Through this work, the team proved that density and thickness anomalies can be used to identify intracranial pressure. From there, Dr. Porras and his team began the work of creating a normative model of bone density and thickness for every age and sex that would give doctors a reference to compare their measurements to. The quickest and most effective method to accomplish this was through artificial intelligence and machine learning.
“During the last few years, we created a normative model — a quantitative reference of cranial development,” Dr. Porras says. “This was essential because we could use our large datasets to learn exactly how a cranium is supposed to grow. That reference has been extremely important because once our AI models learn what is normative, they can very accurately tell when there is any kind of anomaly.”
This normative reference model is publicly available, and the results showing its relevance to identifying intracranial pressure in craniosynostosis are now published. Prospective studies are on the horizon to show the accuracy of this approach in a clinical setting.
With this normative model in place, Dr. Porras hopes that doctors will one day be able to look at patient CT scans and use density and thickness measurements together with artificial intelligence to understand the severity of the pressure and more accurately determine when surgical intervention is needed. This, in turn, could allow children in immediate need to receive earlier treatments while others can wait until they are older to have an invasive surgery.
Citations
- Liu, J., Chaij, J., Linguraru, M.G. et al. Cranial bone thickness and density anomalies quantified from CT images can identify chronic increased intracranial pressure. Neuroradiology 66, 1817–1828 (2024).
- Chaij, Jasmine BS, BA; Liu, Jiawei MS†; French, Brooke MD; Mirsky, David MD; Miles, Randy C. MD, MPH; George Linguraru, Marius DPhil; Nguyen, Phuong D. MD; Alexander, Allyson L. MD, PhD; Görg, Carsten PhD; Porras, Antonio R. PhD. Investigation of Cranial Bone Changes Indicative of Increased Intracranial Pressure in Diverse Phenotypes of Craniosynostosis. Plastic & Reconstructive Surgery-Global Open 13(3):p e6618, March 2025. DOI: 10.1097/GOX.0000000000006618.
- Liu, Jiawei MS; Elkhill, Connor BS; LeBeau, Scott BA; French, Brooke MD; Lepore, Natasha PhD; Linguraru, Marius George DPhil; Porras, Antonio R. PhD. Data-driven Normative Reference of Pediatric Cranial Bone Development. Plastic and Reconstructive Surgery - Global Open 10(8):p e4457, August 2022. | DOI: 10.1097/GOX.0000000000004457.
Featured researcher

Antonio R. Porras, PhD
Research Director
Pediatric Plastic and Reconstructive Surgery
Children's Hospital Colorado
Assistant professor
Biostatistics and Informatics, Colorado School of Public Health
University of Colorado School of Medicine