Project Details

Multimodal Seizure Monitoring in the Neonatal Intensive Care Unit

Felix Richter, MD, PhD

Summary

BACKGROUND: Infant alertness and neurologic changes can signal life-threatening pathology, yet bedside assessments are intermittent and subjective. Reliable, continuous monitoring is needed. GAP: We created and validated NeoPose, a low-cost, non-invasive, computer vision tool to continuously monitor neonates using exclusively video data. NeoPose identifies cerebral dysfunction linked to later neurodevelopmental injury, traditionally identified by electroencephalography and neuroimaging. HYPOTHESIS: We hypothesized that our computer vision method to track movement, NeoPose, could detect neurologic changes in the NICU. METHODS: We collected video-EEG from infants with corrected age less than one year at a level four urban NICU between February 1, 2021 and December 31, 2022. We trained a deep learning pose recognition algorithm on video feeds, labeling 14 anatomic landmarks in 25 frames/infant. We then trained classifiers on anatomic landmarks to predict cerebral dysfunction, diagnosed from EEG readings by an epileptologist, and sedation, defined by the administration of sedative medications. RESULTS: We built the largest video-EEG dataset to date (282,301 video minutes, 115 infants) sampled from a diverse patient population. Infant pose was accurately predicted in three evaluation datasets (ROC-AUCs 0.83–0.94). Median movement increased with age and, after accounting for age, was lower with sedative medications and in infants with cerebral dysfunction (all P<5x10-3, 10,000 permutations). Sedation prediction had high performance (ROC-AUCs 0.87–0.91), as did prediction of cerebral dysfunction (ROC-AUCs 0.76–0.91). IMPACT: We show that NeoPose can be applied in an ICU setting and that an EEG diagnosis, cerebral dysfunction, can be predicted from video data alone. Deep learning on video data may offer a scalable, minimally invasive method for neuro-telemetry in the NICU.