The director of the Neuroinformatics Program at the University of California, Irvine, provided insight on the use of self-organizing maps to cluster different groups of patients with migraine. [WATCH TIME: 4 minutes]
WATCH TIME: 4 minutes
"Now that everyone is working with deep learning models and neural networks, our model is providing a bridge that will connect neural networks to our type of data. The advantage is that it keeps the topographical relationship of variables the same so we can have some sort of visual representation of that data in a 2-dimensional space."
Migraine is defined as a single disorder with several different subtypes; however, there is substantial variability in the clinical manifestations of migraine, both between and within individuals, from attack to attack, and over time. To improve the treatment of migraine, clinicians have begun to systematically map the clinical phenotypes, as well as identify specific treatments for these phenotypes. At the 2023 American Headache Society (AHS) Annual Meeting, held June 15-18, in Austin, Texas, Ali Ezzati, MD, presented research on using a data-driven approach to identify naturally occurring clusters among participants with migraine.
Known as the American Migraine Prevalence and Prevention Study (AMPP), the analysis included a longitudinal, population-based cohort of individuals with who met the modified ICHD-2 criteria for migraine. In the trial, Ezzati and colleagues used a self-organizing map (SOM), which is an unsupervised machine learning technique that is used for dimension reduction, preserves topological features of data, and is based on competitive learning of a 2-layer artificial neural network. Using this method, investigators identified 5 main clusters, comprised of lowest symptom severity, mild symptom severity, moderate severity – high allodynia, high depressive symptoms, and frequent severe migraine.
Ezatti, director of the Neuroinformatics Program at the University of California, Irvine, sat down with NeurologyLive® at the meeting to discuss the research, along with the reasons behind using a SOM. In addition, he provided context on the numerous machine learning techniques making headway throughout clinical research, and the advantages they bring to clinicians.