CDAC Code Catalog

Atrial fibrillation detection during sleep and its relationship to cognitive outcomes.

Code and Data related to aim 'Sleep Staging with AirGo'

AirGo repo for PyT0 version

automating respiration analysis for PSGs

Boston Children's Hospital polysomnography dataset for pediatric sleep research.

This is the repository set up to store all the code related to Step 1 for the BCH sleep pipeline.

Sleep staging models for Boston Children's Hospital polysomnography data.

general code for handling sleep data on BDSP platform.

CAISR - Downloadable/Executable Software Repository

CAISR1+1

sleep analysis powered by AI.

Sleep staging using ECG-derived respiration signals.

Repository for "Sleep in the Intensive Care Unit through the Lens of Breathing and Heart Rate Variability: A Cross-Sectional Study"

Multi-source sleep staging model combining heterogeneous EEG datasets.

ICU-SLEEP (NCT03355053): randomized trial of nocturnal dexmedetomidine in ICU patients — analysis code, deidentification pipeline, and ancillary materials.

Sleep staging analysis for Korean Genome and Epidemiology Study (KoGES) cohort.

noise and light in the ICU analysis

Ordinal regression models for continuous sleep depth estimation from EEG.

Sleep staging optimized for clinical outcome prediction rather than epoch-level accuracy.

PANDA deep learning framework for automated sleep staging across pediatric and adult populations.

The code and trained models that were submitted for the official winning entry of the Physionet 2018: You Snooze You Win challenge

Clinical tools and utilities for sleep medicine practice.

This is the repository to store the code related to EDF to .MAT conversion in the sleep pipeline.

De-identification pipeline for sleep study recordings.

Predicting clinical outcomes from sleep study features and polysomnography data.

Sleep Phenomics Automation

PSG viewer and sleep analysis platform with staging, respiratory, and report views

predict cognitive function from sleep using deep learning

Age-based normative values for sleep EEG parameters across the lifespan.

Code for "Sleep EEG and MRI" project 2022.

General functions for sleep data, reserach etc.

convenient function that format (sleep) biosignals data to common h5 format. used for ICU-Sleep and SleepLab data.

Automated sleep staging model trained and validated on 2,000 patients.

Code and data associated to publication "Refining sleep staging accuracy: Transfer learning coupled with scorability models"

develop multi-task model that can classify stages, apneas, arousals based on respiration signal.

SPiN — Spindle–Oscillation Physiomarkers of Neurohealth

Optimization of sleep spindle detection algorithms and parameters.

Undiagnosed Apnea in the ICU

Reproducibility package for:

Large-scale brain imaging database for multi-modal neurological research.

Brain age estimation for Korean Genome and Epidemiology Study (KoGES) cohort.

Effects of physical exercise on brain age as measured by EEG biomarkers.

Dataset for studying exercise effects on brain health metrics.

Reproducibility package for:

The code repository for the meditation and sleep EEG-based brain age project.

Repo for https://doi.org/10.1093/sleep/zsad294

Transforms overnight sleep EEG into a 1024-D brain-health latent space and an easy-to-use Brain Health Score. Multi-task deep learning predicts cognition, disease risk, and mortality patterns from a single EEG channel.

Brain age estimation from sleep EEG recordings using deep learning.

Computing the breathing stability index from respiratory signals.

Hurst-Ljapunov-Grassberger analysis using Mackey-Glass equations for respiratory signal dynamics.

Predicting acute respiratory instability from CPAP waveform morphology.

Computing self-similarity for high loop gain and central apnea analysis.

Code to automatically detect periods of suppression in critically ill ICU patients.

EEG seizure detection and continuous monitoring analysis tool.

Multi-task learning embeddings for EEG seizure and pattern classification.

This is a repository that will contain all the information regarding the epilepsy project.

Logistic regression models for epilepsy detection from EEG features.

GROND (Generalized Rhythmic and Oscillatory Neurophysiology Descriptor) is an automated system for joint characterization — lateralization, spatial localization, discharge timing, and frequency estimation — of periodic discharges (PD) and…

This is a graphical user interface that allows spike annotations.

Calculates frequency, spatial extent, location, evolution, and other parameters for rhythmic and periodic EEG patterns

Code and dataset used for the development of algorithms that automatically detect the frequency and spatial extent of epileptiform activity, specifically for LPD, GPD, LRDA, GRDA.

Code to reproduce figures in "Interrater Reliability of Expert Electroencephalographers Identifying Seizures and Rhythmic and Periodic Patterns in EEGs"

SPaRCNet deep learning model for classifying IIIC patterns on EEG.

Large-scale mapping and visualization of IIIC patterns across patient populations.

Deep learning framework for EEG pattern classification and seizure detection.

Visualization tool for Morgoth EEG pattern classification results.

Seizure prediction and forecasting model using EEG data.

GUI tool for rapid labeling of IIIC patterns on EEG recordings.

GUI for rapid labeling of segments from multiple EEGs, and instructions for preparing data for these labeling tasks

Real-time EEG monitoring and pattern detection system.

Analysis of seizure forecasting model failures and edge cases.

Pilot clinical trial for validating automated spike detection systems.

Deep learning model for automated interictal epileptiform spike detection (version 1).

Deep learning model for automated interictal epileptiform spike detection (version 2).

code to train and run TEEGLLTEEG

Interactive timeline viewer for EEG events, seizures, and clinical annotations.

NLP-based delirium detection and phenotyping from clinical notes.

Causal survival analysis of delirium progression to dementia.

Code and data to accompany "Physiological Assessment of Delirium Severity: The Electroencephalographic Confusion Assessment Method Severity Score (E-CAM-S)"

EEG-based assessment of Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS).

Inter-rater reliability of EEG-based encephalopathy grading

MATLAB Code for the CAR-T project

Dynamic treatment regimes for ictal-interictal continuum pattern management.

Code and analysis for incorporating inter-rater reliability into EEG feature–based severity modeling, with application to visual EEG grading of encephalopathy.

EEG-based delirium prediction models (Kimchi et al., 2019).

Massachusetts General Hospital nursing Confusion Assessment Method (CAM) dataset.

Multi-modal understanding of sedation and encephalopathy in ICU patients.

RASS-based delirium prediction using EEG features.

VE-CAM-S score for assessing delirium severity from visual EEG features

Code for the VE-ICANS research project

ADARRI: a novel method to detect spurious R-peaks in the electrocardiogram for heart rate variability analysis in the intensive care unit

code to compute CCI and SOFA scores

Variability in Pharmacologically-induced Coma for Treatment of Refractory Status Epilepticus

Code to analyze AirGo and other ICU signals data for Covid-19 patients.

COVID risk prediction code

Multimodal EEG analysis for sedation level monitoring in ICU patients.

ECG, RR intervals, Heart Rate Variability & Co

EEG analysis during therapeutic hypothermia in post-cardiac arrest ICU patients.

Predicting Neurological Outcome in Comatose Patients after Cardiac Arrest with Deep Neural Networks

Simulation framework for ICU EEG pharmacokinetic-pharmacodynamic modeling.

SOFA-LR: a logistic regression based improvement on the SOFA score that accommodates missing data

This is the repository to store the code related to the MOCA project.

Code accompanying the manuscript "Automated Prediction of Glasgow Coma Scale Scores from Unstructured Electronic Health Records: a Natural Language Processing Approach".

Causal matching methods for pharmacokinetic-pharmacodynamic studies in ICU.

Code and data for Ann Neurol . 2018 May;83(5):958-969. doi: 10.1002/ana.25232. Epub 2018 May 16.

The repository contains the implementation for automated prediction of delayed cerebral ischemia after subarachnoid hemorrhage from EEG

computing the Sequential Organ Failure Assessment score

CRIME-PISE algorithm for clinical phenotyping from electronic health records.

EHR-based phenotyping pipeline using MGH/BIDMC notes, ICD codes, and medication records — currently targeting a CHF phenotype.

NAX algorithm for automated brain tumor phenotyping from clinical notes.

NAX algorithm for automated epilepsy phenotyping from clinical notes.

NAX algorithm for mild cognitive impairment and Alzheimer's disease phenotyping.

NAX algorithm for automated narcolepsy phenotyping from clinical notes.

NAX algorithm for automated Parkinson's disease phenotyping from clinical notes.

NAX algorithm for subarachnoid hemorrhage phenotyping from clinical notes.

NAX algorithm for traumatic brain injury phenotyping from clinical notes.

Code for: "A Machine Learning Approach for Identifying People with Neuroinfectious Diseases in Electronic Health Records"

NAX-based phenotyping for subdural hematoma from EHR data.

Code and data to reproduce results in our paper about noise in the diagnosis of epilepsy

A registry of publicly available datasets on AWS

A browser for open-data-registry: https://github.com/awslabs/open-data-registry

Example code for accessing BDSP data on AWS using boto3.

CCNL / BDSP Website

Tools for extracting EMR data for use in BDSP

BDSP open data registry configuration and metadata.

Web application for searching BDSP clinical reports.

Development instance of the bdsp.io web application.

Production instance of the bdsp.io web application.

This is the repository for getting started with connecting with AWS Athena.

This is the repository to store all the AWS Scripts related to BDSP.

Documentation and tools for the Bedmaster ICU project.

This is the repository for the Bedmaster Patient Matching Pipeline.

pipeline - after deidentification portion

This is the repository to store all the code used to deidentify all the BIDMC EEG Reports.

This is the repository to store all the code used to deidentify all the BIDMC Notes.

This is the repository created to share sample code and examples for the Data Science Team to access the BIDMC database.

BIDS-Conversion

Notebook for converting multi-file BDSP .mat EEG recordings into a single EDF+ file and then into a BIDS-EEG dataset.

De-identification of annotations in BIDS-formatted EEG datasets.

Cleanup tool to remove BIDS sessions lacking EDF files.

General-purpose EEG-to-BIDS format conversion tools.

Pipeline for converting EDF files to BIDS format.

code to convert Emory .edfs -> BIDS

Stanford-specific pipeline for EDF-to-BIDS conversion.

Tools for creating and managing BIDS session structures.

Extracting recording start times for BIDS metadata.

Internal development and experimentation workspace.

Downloadable assets (PDFs, slides, checklists, manuscripts) linked from the CDAC / BDSP website

Clinical Data Animations Center (CDAC) web portal for data access.

Haoqi Sun's EEG spike-detection pipeline (ResNet, 10-fold CV with soft labels) trained on the neurological-goog5 GCS dataset.

This is the repository to store the daily updates for the different pipeline that we are using at CDAC.

A code repository designed to show the best GitHub has to offer.

De-identification pipeline for DICOM medical imaging files.

Pipeline for converting discontinuous EDF + recordings to BIDS format, supporting MATLAB and Python implementations.

Pipeline for de-identifying ECG recordings.

Plotting tools for ECG data stored in HDF5 format.

convenience function to write edf from python, e.g. from prepared data

Copy xltek from multiple xltek archives to LM4

This is the repository setup to store all the code related to Step 1 of the EEG Pipeline at BIDMC.

De-identification pipeline for EEG recordings to remove PHI.

This is the repository to store the code related to indexing all the EEG's from MGH and BWH.

Fuzzy matching algorithms for linking EEG records to patient identities.

Tools for parsing and structuring EEG clinical reports.

Tools for normalizing EEG signal data across datasets.

This is the repository to store the code related to the deidentification for Neurology Reports.

Epilepsy.Science

This is a repository to store all the scripts for use cases that might be useful to many people in the CDAC lab.

This is the repository to store all the scripts that we use for our projects.

Grant application management and tracking tools.

Tools for managing and querying the Harvard EEG Database.

This is the repository to store all the code used to deidentify all the MGB Imaging Reports.

inter rater analysis

This is a repository that will contain the code for the lab demo.

Cleaning, standardization, and dose-equivalency tables for MGH medication-administration records (COVID delirium / sedation, ICU sleep studies).

This is the repository to store all the code used to deidentify all the MGB Neurology Reports.

This is the repository to store all the code used to deidentify all the MGB Notes.

This is the code used for converting raw natus files to BIDS format.

Converter for Natus EEG system files to JSON format.

Data processing and analysis orchestration framework.

Batch exporter that turns Persyst P13/P14 .sd4 spike-detection files from MGH LTM EEG recordings into per-recording CSV annotations.

This is the repository to store the code related to deidentifying unstructured clinical data.

Archived BDSP fork of MIT-LCP/physionet-build; active webapp lives in bdsp.io_webapp_prod / bdsp.io_webapp_dev.

Utility for plotting and visualizing ECG signal data.

This is the repository to store the code related to indexing all the Sleep PSG's from MGH.

Real-time streaming data pipeline for continuous EEG monitoring.

Annotator app to label freetext

High-performance data packaging format for large-scale EEG datasets.

Data wrangling utilities for Thunderpack-formatted datasets.

Extraction of annotations from Xltek EEG recording systems.

Multi-agent pipeline for reviewing scientific manuscript figures

Multi-agent pipeline for reviewing and rewriting scientific manuscript drafts before journal submission.

Multi-agent AI pipeline for reviewing scientific manuscripts before journal submission.

ABIEE multi-site data wrangling + Limited Data Set generation

Reader-first Python package for BDSP/Sleep YODA standardized HDF5 data.

This is the code for the BIDMC EEG Archive Pipeline on the Natus Machine.

Clinical trial design search with CHOCOLATE simulation, persisted search history, and Ollama-driven Nemotron guidance.

CICADAS (Causal Inference for Critical-Care Anti-Seizure Treatment with Disease Dynamics, Automated PKPD/Trial Simulation, and Survival Optimization) is a simulation-based framework for emulating randomized trials of anti-seizure treatment…

Causal Inference for Critical-Care Anti-Seizure Treatment with Disease Dynamics, Automated PKPD/Trial Simulation, and Survival Optimization — Modified to accommodate infrequent events and bolus dosing.

Clinical note parsing pipeline for the Stanford CDAC neurology teaching tool. Parses MRI reports, EEG reports, and clinical admission notes into structured JSON, merges them per patient, and synthesizes a single short anatomical phrase for…

No description provided.

Umbrella repo for the PHI receiving pipeline: ingest identified data from many institutions, turn it into an identified OMOP database, then de-identify it into the standardized, controlled-access OMOP research database.

No description provided.

Code accompanying Sartipi et al. — five-stage LLM pipeline for extracting seizure and IIC pattern attributes from free-text EEG reports. Data + DOI on bdsp.io.

EEG Labeling for Rapid Onset/offset Notation and Detection — a PyQt6 tool and machine-learning pipeline for annotating seizure begin/end times on EEG spectrograms at scale, with model-assisted pre-labeling and active learning.

GROND manuscript (TeX + figures); imported into Overleaf. Code lives in bdsp-core/grond.

A new, comprehensive online EEG atlas covering the entire lifespan and every finding present in any existing EEG atlas — with interactive EEG clips, the four-panel qEEG spectrogram, auto-drafted (human-revisable) captions, quiz modes, and…

A growing collection of small, self-contained scripts that automate the kind of tedious, soul-draining tasks academics and researchers keep running into — the things you'd never want to do by hand twice.

icon-diagrams is a code-first figure generation toolkit for publication-style scientific diagrams.

This repository is the home for the design, simulation, and eventually the implementation of the training ("learning with feedback") side of the ILAE EEG skill-certification project. Its sibling, bdsp-core/ilae-skill-certification-test-multi,

Code and data pointers for Kong et al. 2025 (Epilepsia, doi:10.1111/epi.18547) — Evaluating crowdsourcing for ICU EEG annotation

Modeling, simulation, and retrospective feasibility analysis for a short Bayesian-adaptive signal-detection-theory (SDT) test of EEG spike-detection skill, with a three-way sequential stopping rule (pass / fail / refer) calibrated to a…

CORTEX (Continuous Optimization Response Testing for EEG eXpertise): an adaptive Bayesian test of how well a clinician reads an EEG. A joint particle-filter engine over 7 tasks asks the most informative question next and stops once skill is certain. Cross-validated cut scores; Laplace/EKF clinical runtime.

Goal: publish every reproducible paper from the Westover group. (Started as "every JJ first/last-author paper with MBW"; expanded 2026-07-05 to triage all 434 MBW-coauthored papers — see CATEGORIES.md.)

Reproducibility repository for Harid, Jing, Hogan et al., *Measuring Expertise in Identifying Interictal Epileptiform Discharges* (Epileptic Disorders, 2022).

MEEM: EEG↔MRI matching and structured feature extraction (BDSP BIND cohort). Code only; data in s3://bdsp-opendata-credentialed/meem/.

This is the code for running daily updates currently being run on the Megatron Machine using Spark.

This is the pipeline for archiving EEG's from BWH and MGH currently being run on the Megatron Machine.

This is the pipeline for indexing EEG's from BWH and MGH currently being run on the Megatron Machine.

This is the pipeline for indexing PSG's from MGH currently being run on the Megatron Machine.

Normative growth curves (age × sex percentiles, per sleep/wake state) for quantitative EEG slowing features, used to characterize pathological slowing — focal vs. generalized — in the morgoth-viewer automated EEG report pipeline.

This repository extends the base Morgoth EEG model with two improvements specifically targeting seizure detection:

This section describes how to fine-tune the pretrained Morpheus model on your own PSG dataset, using sleep staging as an example.

This repo contains internal Morpheus utilities and outputs. Working directories are organized by module. spindle-so/ is one module/functionality; different team members may own different modules.

A.I. enabled cohort generation tool

Private — CDAC internal. Batch-LLM utilities used across the NAX EHR-phenotyping projects — orchestrates Anthropic / OpenAI batch APIs for large-scale prompting over clinical notes, and provides reusable prompt templates and SOPs.

A unified, continuous, EEG-based index of encephalopathy severity for critical care.

Ultrafast EEG template matching under Dynamic Time Warping — a Python port of the algorithm at the core of NeuroBrowser.

Pipeline for converting clinical Natus/Xltek NeuroWorks EEG recordings into a de-identified, BIDS-compliant dataset and uploading to the Brain Data Science Platform (BDSP).

Private — CDAC internal. Placeholder repository for the KILT (Knowledge Inheritance via Lifelong Training) technical write-up companion to bdsp-core/NoiseinDiagnosing_Epilepsy. This repo currently has no committed code; the analysis lives in the

This repository is for all the OMOP related code for processing MGB parquet files.

Code and documentation for wrangling the BDSP OMOP clinical data — large-scale categorization, classification, and feature extraction over the OMOP CDM (notes, radiology, etc.). Built initially for the BIDMC (site I0002) note-categorization effort.

No description provided.

Planning/design docs for the BDSP research platform (BDC-style) and PHI de-identification pipeline

Analysis code and manuscript for an inter-rater-reliability (IRR) study of sharp transients in EEG: 5,000 segments, each centered on a sharp transient flagged during routine clinical reporting, independently classified by 14…

Reproduction code and data manifest for:

Tools for building the annual Stanford epilepsy division inpatient attending schedule — assigning one EMU attending and one ICU/LTM EEG attending to each of the year's 52 weeks, honoring everyone's availability, requested week counts, and preferences.

Code to wrangle the teleEEG (Neurotech) EEG dataset published on bdsp.io: run the Morgoth foundation-model prediction heads over every recording, compute performance vs.

Private — CDAC internal. Data-wrangling notebooks fixing common issues in the VXP (clinical reports) dataset — invalid rows, missing OrderIDs, malformed columns — plus an LLM-assisted code fixer used as part of the cleanup.