Find resources for your research

A credentialed, harmonized commons of de-identified clinical sleep and neurophysiology data: polysomnography, EEG, brain MRI and linked EHR, from multiple health systems, developed under the ARPA-H REST program.

Pediatric SleepNet: a multimodal (EEG+EOG+EMG) U-Net deep learning model for automated pediatric sleep staging across development (infants to adolescents), with age-specific models trained on the BCH Sleep Corpus. Trained model weights + code.

LENS: lifespan and sleep-stage-resolved normative EEG growth charts that score background slowing as a deviation from age- and stage-matched normal, for detection and automated reporting. De-identified derived data and reproducible code.

Automated characterization of periodic discharges and rhythmic delta activity in critical-care EEG: frequency, laterality, spatial extent and individual discharge timing, benchmarked against four expert raters.

Multicenter clinical and EEG Database including patients who were comatose after cardiac arrest. This database includes 1,020 patients and more than 50,000 hours of continuous EEG from 7 hospitals in the U.S. and Europe.

ELUCID is a 7-center longitudinal study of 600 adults aged 55+ with late-onset unexplained epilepsy, building a multimodal biorepository (brain MRI, overnight EEG, cognitive testing, blood biomarkers) to identify risk factors for dementia.

De-identified data and a PK-PD + MALTS causal-inference pipeline reproducing the effect of untreated epileptiform-activity burden on poor discharge outcome in critically ill patients (Parikh et al., Lancet Digital Health 2023).

MATLAB code and de-identified data (1,991 patients) relating automatically-quantified ictal-interictal-continuum/seizure burden on continuous EEG to discharge outcome (Zafar et al., Ann Neurol 2021).

MATLAB/Python code and de-identified data for a simulation framework that estimates required sample sizes and effect sizes for randomized trials of anti-seizure treatment in critically ill patients (Parikh et al., Ann Clin Transl Neurol 2024).

De-identified sleep-spindle feature data (BDSP-linked) + cognition data + Python code to find the spindle-detection parameters that best predict cognitive performance (Adra et al., Sleep 2022).