Resources for BMIF204

MIMIC-IV Setup Scripts

wget https://raw.githubusercontent.com/apvidul/setup-scripts/refs/heads/main/setup_ehr_tutorial_mimic4_codi_nlp.sh

Tools for Downstream Analysis

1. ONCE: A feature generation app that identifies related NLP and codified features based on an input or target disease. For example: PheCodes and CUIs.

2. KESER Network: A tool to identify and visualize codified concepts relevant to diseases, medications, and procedures (e.g., PheCodes, RxNorm).

3. NILE: An NLP tool for fast and efficient processing of clinical notes to UMLS CUIs. Note: You will need a UMLS License to use this tool.

4. PETEHR: Python toolkit for EHR processing developed specifically for this tutorial. It is designed to process MIMIC-IV data and generate research-ready datasets.

 

Accessing Jupyter Notebook on O2 from your local machine

Follow the steps below to set up and access a Jupyter Notebook running on the o2 server from your local machine.

1. Prerequisites

Ensure that X11 forwarding is active on your system:

• Mac: Install and run XQuartz.
• Windows: Install MobaXterm.

2. Set Up Local Port Forwarding

On your local machine, run the following command to log in to the o2 server and set up port forwarding:

ssh -Y -L 50001:127.0.0.1:50001 username@o2.hms.harvard.edu

3. Request an Interactive Session

After logging into o2, request an interactive compute node enabling X11 forwarding and setting up a network tunnelby running:

srun -t 0-00:40 --pty -p interactive --mem=12G --x11 --tunnel 50001:50001 /bin/bash

4. Navigate to Your Workspace and Activate Conda

cd /n/scratch/users/v/va67/EHR_TUTORIAL_WORKSPACE
source activate ehr_tutorial

5. Start the Jupyter Notebook Server on O2

jupyter notebook --port=50001 --browser="none"

6. Access the Jupyter Notebook on your Local Machine

Open a browser and go to http://localhost:50001/tree

7. Shut Down the Jupyter Server

To stop the Jupyter Notebook server, go to the terminal where the session is running and do Ctrl+C