Resources

MATLAB code and de-identified data to forecast the onset and day-by-day trajectory of ICANS after CAR T-cell therapy (combined hidden Markov model + lasso logistic regression; 199-patient cohort, leave-one-patient-out cross-validation).

A continuous, EEG-based index of encephalopathy severity (NESI) that maps RASS, GCS, CAM-S, and ICANS onto one physiological axis. Includes 10-minute EEG segments, MORGOTH features, trained models, derived tables, and the full analysis code.

Deidentified seizure and ictal-interictal continuum (IIC) data extracted from 156,582 EEG reports across three health systems (two adult, one pediatric), plus the LLM-based extraction pipeline. Companion to Sartipi et al. (IJMI, under review).

Dataset and code for developing and validating machine learning models to phenotype narcolepsy type 1 (NT1) and narcolepsy type 2/idiopathic hypersomnia (NT2/IH) from multi-site electronic health record data, including cross-sectional classification

Using sleep EEG to predict future health outcomes.

a MCI/ADRD EHR phenotyping model trained with python sklearn pipeline, injoblib format.

An expert-level deep learning model for detecting epileptiform discharges in EEG recordings at both event and whole-EEG levels.

Data and code to reproduce this paper: https://pmc.ncbi.nlm.nih.gov/articles/PMC9847346/

TEEGLLTEEG

Algorithm to detect breathing pattern self-similarity to predict CPAP therapy outcomes in OSA