Project Grant R21NS151859

The National Institute of Neurological Disorders and Stroke awarded the University of Alabama at Birmingham $186,250 on September 3, 2026, under the Extramural Research Programs in the Neurosciences and Neurological Disorders program (CFDA 93.853) to develop and validate machine-learning seizure risk prediction models for newborns.

The project, titled Prediction of Risk of Infant Seizures Using Machine Learning (PRISM), addresses a critical gap in neonatal seizure detection. Existing automated seizure identification models focus on identifying seizures after onset; this award funds development of reliable, clinically interpretable models to predict seizure risk in real time using short-duration electroencephalography (EEG) recordings of one hour or less, integrated with multimodal data including video and clinical features. The hypothesis underlying the work posits that seizure predictive biomarkers within short-duration EEGs will enable discrimination of seizure onset within 10 to 23 hours, achieving a receiver operating characteristic area under the curve of 0.80 or greater. Preliminary feasibility work on 20 patients demonstrated promising results, with a decision tree classifier achieving an area under the curve of 0.78, supporting progression to full validation.

Performance of work occurs in Birmingham, Alabama. The period of performance runs from September 3, 2026, through August 31, 2028. The assistance type is a Project Grant, classified as an R21 exploratory/developmental research award.