NiChart: Neuro-imaging Chart

NiChart is a comprehensive framework designed to revolutionize neuroimaging research. It offers large-scale neuroimaging capabilities, sophisticated analysis methods, and user-friendly tools, all seamlessly integrated into a local installation version and the AWS Cloud.

Components

1. Image Processing: Utilizes tools like DLMUSE, fMRIPrep XCEPengine, and QSIPrep for effective image analytics.
2. Reference Data Curation: Houses ISTAGING, 70000 Scans, and 14 individual studies to provide curated reference data.
3. Data Harmonization: Employs neuroharmonize and Combat for ensuring consistent data standards.
4. Machine Learning Models: Provides Supervised, Semi-supervised, and DL Models for advanced neuroimaging analysis including SpareScore.
5. Data Visualization: Features like Centile curves, direct image linking, and reference values for comprehensive data visualization.
6. Deployment: Supports open-source Github components and Docker container compatibility deployed in a local environment & AWS Cloud.

System Requirements

For recommended system configuration, please refer to: nnUNet hardware requirements.

Installation Instructions

1. (Optional but recommended for environment management) Mamba installation Mamba Installation Guide (Official)

   Example (Linux x86):

   wget https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-Linux-x86_64.sh
   
   bash Mambaforge-Linux-x86_64.sh
   mamba create -c conda-forge -c bioconda -n NCP_env python=3.12 
   mamba activate NCP_env

2. Install NiChart_Project into the environment

   git clone https://github.com/CBICA/NiChart_Project.git
   pip install -r requirements.txt

3. Install the proper PyTorch version for your device Numpy and PyTorch have some compatibility issues which have changed variously on different platforms. To avoid frustration with these issues, please install PyTorch as noted below.

   After installing all other requirements, uninstall Torch:

   pip uninstall torch
   

   Then install PyTorch using the following command. Make sure to use the correct index url for your CUDA version as specified on the PyTorch getting started page. On Linux, use version 2.3.1. On Windows, 2.5.1 is known to work.

   pip install torch==2.3.1 --index-url https://download.pytorch.org/whl/cu121
   

Run NiChart Locally (GUI)

cd src/viewer/
streamlit run NiChartProject.py

The app will start in your localhost.

Quick Links

© 2024 CBICA. All Rights Reserved.