Researchers have leveraged artificial intelligence to identify three sub-types of Chiari type-1 malformation, which will assist medical professionals in making better treatment choices for their patients. Chiari type-1 malformation occurs when part of the brain, specifically the cerebellum, extends beyond the skull into the spinal canal.
Approximately 4% of the population experiences this congenital brain malformation, which has continually posed challenges in understanding its causes and treatments. The diagnosis for Chiari type-1 malformation is relatively straightforward: the cerebellum protrudes at least five millimeters through the opening in the skull connecting to the spinal cord.
There is no single known cause for Chiari type-1, and the associated symptoms can vary significantly. Common issues include chronic headaches, difficulty swallowing, decreased muscle strength, and syringomyelia, a condition characterized by cyst formation in the spinal cord. Symptoms can appear alone or in combination, and many individuals live their lives without experiencing any significant issues. This variability has complicated the development of effective treatment protocols by healthcare providers.
A study by researchers at Washington University in St. Louis aims to bridge this gap. Through a collaboration of neurosurgeons and computer scientists, they have categorized three distinct sub-types of Chiari type-1 that can help doctors tailor treatment options.
The findings have been published in the journal Neurosurgery.
More comprehensive information has become crucial to refine Chiari type-1 diagnoses, enabling physicians to reliably identify which cases may require clinical intervention. For example, surgical procedures can expand the opening at the skull’s base to alleviate pressure on the brain and improve symptoms for some Chiari type-1 patients.
“Chiari accounts for a significant portion of the cases pediatric neurosurgeons encounter—it’s likely one of the top three reasons for surgeries,” remarked Sean Gupta, MD, a neurosurgery resident at WashU Medicine and co-lead author of the study. He noted that not all patients are candidates for, or benefit from, surgery. The procedure is particularly important for those with both syringomyelia and headaches, while some patients may find relief through pain management or observation. Additionally, many cases go unnoticed by both patients and their physicians.
“In certain studies surveying randomly selected individuals who underwent MRIs but received no diagnosis, we found that up to 4% exhibit Chiari malformations, yet they may not experience any related problems,” Gupta explained.
For patients whose health and quality of life are affected, doctors previously lacked comprehensive information on how to address a variety of symptoms that did not consistently respond to treatment.
There was ample data available to analyze symptom patterns and malformations to identify subtypes for effective treatment protocols. WashU Medicine leads the Park-Reeves Syringomyelia Research Consortium, which entails data from over 1,200 Chiari type-1 patients analyzed for correlations. Each patient has numerous variables to consider—ranging from clinical information and brain imaging to health insurance details. A subset of these variables was meticulously chosen using a mix of data-driven techniques and clinician feedback acquired through surveys of pediatric neurosurgeon experts across the nation.
“This is what we refer to as a very high-dimensional problem, as there are multiple variables at play,” stated Chenyang Lu, PhD, a co-senior author of the study and Fullgraf Professor of Computer Science & Engineering at the McKelvey School of Engineering and founding director of the AI for Health Institute at WashU. Co-senior author David Limbrick, MD, PhD, who now holds the James W. and Frances G. McGlothlin chair of the Department of Neurosurgery at Virginia Commonwealth University School of Medicine, sought Lu’s expertise because artificial intelligence tools excel at analyzing extensive datasets to uncover patterns and correlations related to Chiari type-1 cases. These patterns may provide guidance for physicians in determining the optimal treatment strategies for their patients.
Ziqi Xu, a PhD student in Lu’s lab at WashU and co-lead author, developed the AI algorithm to examine over 500 variables within the dataset, leading to the identification of three unique subtypes of Chiari type-1.
Patients in Cluster 1 were predominantly female, generally diagnosed later in childhood, and reported chronic headaches along with few other health complications. Cluster 2 comprised younger patients who experienced fewer headaches but exhibited a broader range of issues such as difficulties with muscle control and swallowing. The third cluster often presented with spinal deformities, potentially warranting standard decompression surgery and further spinal procedures.
“This research should aid in creating guidelines to determine which patients require surgery, the type of surgery needed, or what alternative therapies may be appropriate,” Gupta noted. “We need evidence-based consensus to inform clinicians on how to manage these patients, as we have been operating with quite incomplete data up to this point.”
Xu is currently working to further enhance and refine this model, expressing her belief that the teamwork between clinicians and computer scientists in this study could revolutionize medical practice.
“We are experiencing an exciting era in medicine,” Xu added. “As computational tools become more advanced and electronic health records grow in size, AI can catalyze new insights for clinicians, allowing us to collaborate effectively towards impactful discoveries and improved patient care.”
