B Tool Installation

title: B Tool Installation
description: Introduces installation methods for common tools.

B Tool Installation

1. Introduction

Medical imaging algorithm development typically relies on the collaboration of multiple toolchains. This chapter will introduce the core environments required for this tutorial, including Python, ASTRA, BART, MONAI, NiBabel and other basic components. We will start with basic Python environment configuration and gradually explain how to install CT reconstruction library (ASTRA), MRI reconstruction toolbox (BART), deep learning framework (MONAI), and medical imaging file reading/writing tools (NiBabel). Through this section, readers can quickly build a complete, reproducible medical imaging development and reconstruction experimental environment.

2. Direct Python Installation

2.1. Windows System

2.1.1. Download Python Installer

Recommended method: Download the installer from the official Python website

Open your browser and visit: https://www.python.org
Click Downloads in the top navigation bar
The page will automatically recommend the version suitable for Windows, for example: Download Python 3.xx
Click this button to download a .exe installer (e.g., python-3.12.3-amd64.exe)

2.1.2. Run the Installer (Most Critical Step)

Double-click the downloaded python-3.xx-amd64.exe and follow these steps:

At the bottom of the installation interface, there is a line:
- ✅ Must check: Add Python 3.xx to PATH
- This step determines whether you can directly use the python command in the command line later, which is very important.
In Customize installation:
1. Optional Features It is recommended to check all:
  - ✅ pip (Python package manager, essential)
  - ✅ IDLE (built-in small editor for visual debugging of simple scripts)
  - ✅ Documentation
  - ✅ Python test suite
  - ✅ py launcher Then click Next
2. Advanced Options: It is recommended to check:
  - ✅ Install for all users (recommended for multi-user computers, path will be in C:\Program Files\Python3x)
  - ✅ Add Python to environment variables (if you missed checking PATH earlier, you must check it here)
  - ✅ Precompile standard library (speeds up first run)

2.1.3. Verify Python Installation

After installation is complete, follow these steps to check:

Press the Win key, type cmd, and open Command Prompt
In the black window, type:
bash
```
python --version
```
1
Or
bash
```
py --version
```
1
If you see something like:
text
```
Python 3.12.3
```
1
It means the installation was successful and PATH is working.

2.1.4. Verify pip Availability

In the same command line, type:

bash

pip --version

If you see something like:

text

pip 24.x from C:\Python\Python312\Lib\site-packages\pip (python 3.12)

It means pip has been successfully installed and can be used to install subsequent packages like monai and nibabel.

2.2. Linux System

2.2.1. Check if Python is Already Installed

Open the terminal and type:

bash

python3 --version

If you see output similar to:

text

Python 3.10.12

It means the system already has Python3, which can generally be used directly.

Then check if pip exists:

bash

pip3 --version

If it prompts "command not found", it means pip is not installed yet and will be installed later.

2.2.2. Install Python Using Package Manager (Ubuntu / Debian)

2.1 Update Software Sources

First update the package list:

bash

sudo apt update

2.2 Install Python3 and Common Dependencies

Execute:

bash

sudo apt install -y python3 python3-pip python3-venv

Description:

python3: Python interpreter
python3-pip: Python package management tool pip
python3-venv: Used to create virtual environments (very useful when not installing Conda)

After installation, confirm the versions again:

bash

python3 --version
pip3 --version

3. Install Python Using Conda (Recommended)

Conda is the most commonly used environment management tool in the field of data science and medical imaging, capable of creating independent Python environments for different projects to avoid package conflicts.
This section will introduce how to install Conda (Miniconda) on Windows / Linux and create a clean, controllable Python environment.

3.1. Download and Install Miniconda (Recommended)

Why use Miniconda instead of Anaconda?

Miniconda is more lightweight and doesn't install thousands of unnecessary packages
More suitable for scientific research projects and medical imaging engineering
Fully compatible with Anaconda's functionality

3.2. Windows Miniconda Installation

Step 1: Download Installer

Visit the official website:

👉 https://repo.anaconda.com/miniconda/

Select:

Miniconda3 Windows 64-bit Installer (.exe)

Step 2: Run Installer

Double-click the installer:

Choose Just Me or All Users
Important: Check "Add Miniconda3 to my PATH environment variable"
- If you don't check this, it's okay, Conda will automatically add PATH for CMD / PowerShell
Click Install

Step 3: Verify Installation

Open Anaconda Prompt or CMD:

bash

conda --version

If it displays:

conda 24.x.x

Then the installation is successful.

3. Linux Miniconda Installation

Step 1: Download Installation Script

Visit the official website:

👉 https://docs.conda.io/en/latest/miniconda.html

Copy the Linux installation script link, for example:

bash

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Step 2: Run Installation Script

bash

bash Miniconda3-latest-Linux-x86_64.sh

Follow the prompts:

Read and agree to the license
Choose installation path (default is fine: ~/miniconda3)
Choose whether to initialize conda (recommended YES)

Step 3: Activate Conda

If you chose YES earlier, just reopen the terminal.
If you chose NO, you need to manually run:

bash

source ~/miniconda3/bin/activate

Verify:

bash

conda --version

Step 4: Create a New Python Environment (Most Critical)

Whether on Windows / Linux, the following steps are completely consistent.

Create a dedicated environment for medical imaging projects, for example medimg:

bash

conda create -n medimg python=3.10

Explanation:
-n medimg: Environment name
python=3.10: Specify Python version (most compatible with MONAI/ASTRA/BART)

Activate the environment:

bash

conda activate medimg

Verify:

bash

python --version

You should see:

Python 3.10.x

4. Install ASTRA

4.1. ASTRA Toolbox Introduction

ASTRA Toolbox Official Links:
Official website: https://astra-toolbox.com/
GitHub repository: https://github.com/astra-toolbox/astra-toolbox

ASTRA Toolbox (All Scale Tomographic Reconstruction Antwerp) is a high-performance computing library specifically designed for X-ray tomography (CT). It provides GPU-accelerated projection and reconstruction algorithms, making it one of the preferred tools for researchers to perform:

Parallel beam / fan beam / cone beam CT geometric simulation
Forward projection
FBP (Filtered Back Projection) fast reconstruction
Iterative reconstruction algorithms such as ART, SIRT, CGLS
Deep learning + CT simulation / reconstruction preprocessing

4.2. Windows System ASTRA Toolbox Installation

Install via Conda (Most Recommended)

This is the preferred method for Windows: simple, stable, and no compilation required.

Step 1: Create Python Environment

It is recommended to first create a clean Conda environment (to avoid contaminating the system Python):

bash

conda create -n medimg python=3.10
conda activate medimg

Python 3.9–3.11 is recommended (ASTRA official support range).

Step 2: Install ASTRA (CPU Version)

Windows does not support GPU version, so use directly:

bash

conda install -c astra-toolbox astra-toolbox

Conda will automatically install:

astra-toolbox (Python interface)
astra-core (C++ core library)
Related dependencies (such as numpy, scipy)

Step 3: Verify Installation

Enter Python:

bash

python

Type:

python

import astra
print(astra.__version__)

If the version number (such as 1.10.0) is output normally, the installation is successful.

4.2. Linux System ASTRA Toolbox Installation

Step 1: Create Environment (Strongly Recommended Independent Environment)

bash

conda create -n medimg python=3.10
conda activate medimg

Step 2: Install ASTRA (Automatically Select CPU/GPU)

Execute the officially recommended command:

bash

conda install -c astra-toolbox -c nvidia astra-toolbox

Description:

-c astra-toolbox: ASTRA official repository
-c nvidia: Used to provide CUDA runtime (required for Linux GPU)
If your system does not have a GPU, the CPU version will be installed
If you have a GPU, the GPU-accelerated version will be installed (CUDA runtime automatically installed)

No need to manually install CUDA toolkit!

Step 3: Test Installation

bash

python - << 'EOF'
import astra
print("ASTRA version:", astra.__version__)
EOF

If the version number is output, it's successful.

4.3. Test Installation Results

You can use this code to verify if ASTRA is working properly:

python

import astra
import numpy as np

# 64x64 phantom
vol = np.ones((64, 64), dtype=np.float32)
vol_geom = astra.create_vol_geom(64, 64)

# Parallel beam geometry (180 angles)
angles = np.linspace(0, np.pi, 180, endpoint=False)
proj_geom = astra.create_proj_geom('parallel', 1.0, 64, angles)

pid, sino = astra.create_sino(vol, proj_geom)
print("sino shape:", sino.shape)

Expected output:

sino shape: (180, 64)

5. Install BART

5.1. BART Introduction

Official website: https://mrirecon.github.io/bart/
GitHub repository: https://github.com/mrirecon/bart
Official documentation: https://bart-doc.readthedocs.io/en/latest/intro.html
Workshop and example material repository: https://github.com/mrirecon/bart-workshop

BART (Berkeley Advanced Reconstruction Toolbox) is an open-source toolbox for MRI (Magnetic Resonance Imaging) reconstruction, signal processing, and rapid prototyping development. It was developed by UC Berkeley and is one of the most widely used reconstruction tools in the medical imaging and MR physics community, especially suitable for researchers.

BART is a high-performance MRI reconstruction and signal processing toolkit characterized by:

Command-line tools + C library + Python/MATLAB interfaces
Supports basic MRI reconstruction:
- FFT / IFFT
- Sense / pSense
- GRAPPA
Supports advanced algorithms:
- CS (Compressed Sensing)
- L1-wavelet / TV regularization
- LLR, LORAKS, low-rank reconstruction
- NUFFT (Non-uniform FFT)
Supports MRI acquisition geometries:
- Multi-coil data
- k-space non-uniform sampling
- Multi-dimensional data (2D/3D/dynamic MRI)

5.2. Windows System BART Installation

Step 1: Environment Preparation

Windows system.
Install a compatible GLC compiler environment (BART has limited official support on Windows). The official provides two paths: use Cygwin or run Linux in a virtual machine.
In the Cygwin environment, you need to install the following packages:
- Devel: gcc, make
- Math: fftw3, fftw3-doc, libfftw3-devel, libfftw3_3
- Math: liblapack-devel, liblapack-doc, liblapack0
If you want GPU acceleration (Windows itself has weak support)—it's generally recommended to do this on Linux. The documentation points out that "Running BART on Windows is not supported" but some users run it through Cygwin/WSL.

Step 2: Download BART Source Code or Release Package

Open the repository: https://github.com/mrirecon/bart
Or the official webpage: https://mrirecon.github.io/bart/installation.html → Download the latest version zip/tar package.
Download the latest Release (such as version 0.9.00) with Windows support instructions.
Extract the compressed file to a directory, for example C:\tools\bart-0.9.00\

Step 3: Compile and Install Using Cygwin

Install Cygwin: Visit https://www.cygwin.com/, download the installer.
When installing Cygwin, select the corresponding packages in the installation interface (see "Environment Preparation").
Open Cygwin Terminal, in the extracted BART directory, run:
bash
```
cd /cygdrive/c/tools/bart-0.9.00
make
```
1
2
This will compile BART's command-line tools and libraries.
If you want to support GPU (depends on whether there's an adaptation on Windows, it may fail), you can try adding CUDA=1 in the Makefile, but the official warns that Windows support is limited.

Step 4: Verify Installation on Windows (Cygwin)

In the Cygwin shell, try running a BART tool command:

bash

bart fft   # If this command outputs help or an error prompt "usage", then installation is successful

When calling in Python / MATLAB (if you compiled the Python interface), ensure the library path is in the environment variables.
If you encounter "command not found" or library loading failure, you can check if the environment variable PATH contains BART's bin directory, or if Cygwin's usr/bin has been linked.

Tips and Common Issues

BART support is low in Windows environments, strongly recommend using WSL2 or Linux virtual machines for more stable operation. The documentation explicitly states statements like "Windows support by MSYS2; generic" or "Use Cygwin".
If just for algorithm reproduction experiments, you can also enable WSL2 on Windows to install Ubuntu, then follow the Linux installation process within WSL2.
Common errors during compilation: missing fftw3, lapack, blas libraries, ensure you select the corresponding dev packages when installing Cygwin.
GPU support is extremely unstable on Windows and not recommended for early research stages.

5.3. Linux System BART Installation

Step 1: Environment Preparation

Recommended dependencies to install first:

bash

sudo apt-get update
sudo apt-get install -y \
    gcc \
    make \
    libfftw3-dev \
    liblapacke-dev \
    libpng-dev \
    libopenblas-dev

gcc, make: Compilation tools.
libfftw3-dev: FFTW (Fast Fourier Transform Library) development package.
liblapacke-dev: LAPACK/BLAS related development package.
libpng-dev: If you need image reading/writing support.
libopenblas-dev: Recommended to accelerate BLAS operations.

Step 2: Download BART Source Code or Release Package

Download the latest version from the official page or GitHub repository:

bash

git clone https://github.com/mrirecon/bart.git
cd bart

Or download .tar.gz from the release page:

bash

wget https://github.com/mrirecon/bart/archive/v0.9.00.tar.gz
tar xzvf v0.9.00.tar.gz
cd bart-0.9.00

Step 3: Compile Install, Enable GPU Acceleration (Optional)

In the source code directory, run:

bash

make

If you want to enable additional features (such as ISMRMRD support):

bash

make ismrmrd

If your machine has an NVIDIA graphics card and you want to use GPU acceleration, BART supports CUDA.

Execute in the source code directory during compilation:

bash

make clean
make CUDA=1

This enables GPU support. You need to install the corresponding CUDA version, NVIDIA driver, CUDA Toolkit, etc.

Step 4: Install Python Interface (Optional)

If you plan to use BART in Python, many versions have a built-in python/ directory in the source code. It is recommended:

bash

cd python
pip install .

Then test in Python:

python

import bart
bart.print_settings()

Step 5: Verify Installation is Successful

Execute in the terminal:

bash

bart fft   # Display help information

You can also run a simple command-line reconstruction tool to test. If it can output the version number, help information, or no errors, it indicates successful installation.

6. Install MONAI

6.1. MONAI Introduction

Official website: https://monai.dev/
GitHub repository: https://github.com/Project-MONAI/MONAI
Official documentation: https://docs.monai.io/ (including installation guide, API documentation, etc.)

MONAI (Medical Open Network for AI) is a medical imaging deep learning framework jointly led by NVIDIA + US NIH (National Institutes of Health).
It is based on PyTorch and specifically designed for medical imaging tasks (CT / MRI / X-ray / Ultrasound), providing a complete toolchain from data loading, preprocessing, training, evaluation to deployment.

MONAI has become one of the most mainstream open-source frameworks in medical imaging AI research.

6.2. Pre-Installation Preparation

Before installing MONAI, you need to:

Install Miniconda or Anaconda
Install Python 3.9–3.11 (3.10 recommended)
Optional: Install NVIDIA drivers (if GPU is needed)
It is recommended to use an independent conda environment

6.3. Install PyTorch

MONAI depends on PyTorch, so PyTorch must be installed first.

Go to the PyTorch official installation page:

👉 https://pytorch.org/

Select the corresponding system's command.

● CPU Version (Windows / Linux Universal)

bash

pip install torch torchvision torchaudio

● GPU Version (using CUDA 12.4 as an example)

bash

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

If you don't know the CUDA version, run nvidia-smi (Linux) or check the NVIDIA Control Panel (Windows).

6.4. Install MONAI

bash

pip install monai

Install Optional Extension Dependencies (Optional)

Common enhancement packages:

bash

pip install "monai[nibabel,skimage,pillow,ignite]"

All optional dependencies (most complete version):

bash

pip install "monai[all]"

Suitable for research tasks (such as segmentation / registration).

6.5. Installation Verification

Run the following Python script:

python

import monai
import torch

print("MONAI version:", monai.__version__)
print("PyTorch version:", torch.__version__)
print("CUDA available:", torch.cuda.is_available())

If there are no errors and the existing versions are output, it means the installation is successful.

7. Install NiBabel

7.1. NiBabel Introduction

GitHub repository (source code + issues + developers): https://github.com/nipy/nibabel
Official documentation: https://nipy.org/nibabel/
PyPI: https://pypi.org/project/nibabel/

NiBabel is a Python toolkit specifically designed for reading, processing, and saving medical imaging files, supporting mainstream formats such as NIfTI (.nii/.nii.gz), Analyze, MINC, MGH/MGZ, etc. It can load medical imaging data as NumPy arrays while providing complete metadata management such as affine matrices, spatial orientations, and header information. It is one of the most commonly used file I/O libraries in medical imaging AI, neuroimaging (fMRI/dMRI), deep learning preprocessing, and scientific analysis, and is also one of the underlying basic components of MONAI, PyTorch, and various medical imaging toolchains.

7.2. Install Current Stable Version (Recommended)

Use pip to install the latest released version of NiBabel:

bash

pip install nibabel

7.3. Install Latest Development Version

If you want to use the latest development progress on GitHub (unreleased version), you can run:

bash

pip install git+https://github.com/nipy/nibabel

7.4. Install in "Editable Mode" (Used when Developing NiBabel Source Code)

When you need to modify NiBabel source code, participate in development, or debug, you can install in editable mode:

bash

git clone https://github.com/nipy/nibabel.git
pip install -e ./nibabel

This method makes Python directly reference the local source code directory, so no reinstallation is needed after modifications.

7.5. Test NiBabel

1. Run Complete Tests During Development (Recommended for Developers Using tox)

bash

git clone https://github.com/nipy/nibabel.git
cd nibabel
tox

tox will automatically create virtual environments and test NiBabel under multiple Python versions, suitable for users participating in development.

2. Test Installed NiBabel

If you just want to test the NiBabel installed in your current system, you can install test dependencies and run pytest:

bash

pip install nibabel[test]
pytest --pyargs nibabel

This will run all test modules of NiBabel to ensure the installation is normal.

B Tool Installation ​

B Tool Installation ​

1. Introduction ​

2. Direct Python Installation ​

2.1. Windows System ​

2.1.1. Download Python Installer ​

2.1.2. Run the Installer (Most Critical Step) ​

2.1.3. Verify Python Installation ​

2.1.4. Verify pip Availability ​

2.2. Linux System ​

2.2.1. Check if Python is Already Installed ​

2.2.2. Install Python Using Package Manager (Ubuntu / Debian) ​

2.1 Update Software Sources ​

2.2 Install Python3 and Common Dependencies ​

3. Install Python Using Conda (Recommended) ​

3.1. Download and Install Miniconda (Recommended) ​

3.2. Windows Miniconda Installation ​

Step 1: Download Installer ​

Step 2: Run Installer ​

Step 3: Verify Installation ​

3. Linux Miniconda Installation ​

Step 1: Download Installation Script ​

Step 2: Run Installation Script ​

Step 3: Activate Conda ​

Step 4: Create a New Python Environment (Most Critical) ​

4. Install ASTRA ​

4.1. ASTRA Toolbox Introduction ​

4.2. Windows System ASTRA Toolbox Installation ​

Install via Conda (Most Recommended) ​

Step 1: Create Python Environment ​

Step 2: Install ASTRA (CPU Version) ​

Step 3: Verify Installation ​

4.2. Linux System ASTRA Toolbox Installation ​

Step 1: Create Environment (Strongly Recommended Independent Environment) ​

Step 2: Install ASTRA (Automatically Select CPU/GPU) ​

Step 3: Test Installation ​

4.3. Test Installation Results ​

5. Install BART ​

5.1. BART Introduction ​

5.2. Windows System BART Installation ​

Step 1: Environment Preparation ​

Step 2: Download BART Source Code or Release Package ​

Step 3: Compile and Install Using Cygwin ​

Step 4: Verify Installation on Windows (Cygwin) ​

Tips and Common Issues ​

5.3. Linux System BART Installation ​

Step 1: Environment Preparation ​

Step 2: Download BART Source Code or Release Package ​

Step 3: Compile Install, Enable GPU Acceleration (Optional) ​

Step 4: Install Python Interface (Optional) ​

Step 5: Verify Installation is Successful ​

6. Install MONAI ​

6.1. MONAI Introduction ​

6.2. Pre-Installation Preparation ​

6.3. Install PyTorch ​

6.4. Install MONAI ​

6.5. Installation Verification ​

7. Install NiBabel ​

7.1. NiBabel Introduction ​

7.2. Install Current Stable Version (Recommended) ​

7.3. Install Latest Development Version ​

7.4. Install in "Editable Mode" (Used when Developing NiBabel Source Code) ​

7.5. Test NiBabel ​

1. Run Complete Tests During Development (Recommended for Developers Using tox) ​

2. Test Installed NiBabel ​

B Tool Installation

B Tool Installation

1. Introduction

2. Direct Python Installation

2.1. Windows System

2.1.1. Download Python Installer

2.1.2. Run the Installer (Most Critical Step)

2.1.3. Verify Python Installation

2.1.4. Verify pip Availability

2.2. Linux System

2.2.1. Check if Python is Already Installed

2.2.2. Install Python Using Package Manager (Ubuntu / Debian)

2.1 Update Software Sources

2.2 Install Python3 and Common Dependencies

3. Install Python Using Conda (Recommended)

3.1. Download and Install Miniconda (Recommended)

3.2. Windows Miniconda Installation

Step 1: Download Installer

Step 2: Run Installer

Step 3: Verify Installation

3. Linux Miniconda Installation

Step 1: Download Installation Script

Step 2: Run Installation Script

Step 3: Activate Conda

Step 4: Create a New Python Environment (Most Critical)

4. Install ASTRA

4.1. ASTRA Toolbox Introduction

4.2. Windows System ASTRA Toolbox Installation

Install via Conda (Most Recommended)

Step 1: Create Python Environment

Step 2: Install ASTRA (CPU Version)

Step 3: Verify Installation

4.2. Linux System ASTRA Toolbox Installation

Step 1: Create Environment (Strongly Recommended Independent Environment)

Step 2: Install ASTRA (Automatically Select CPU/GPU)

Step 3: Test Installation

4.3. Test Installation Results

5. Install BART

5.1. BART Introduction

5.2. Windows System BART Installation

Step 1: Environment Preparation

Step 2: Download BART Source Code or Release Package

Step 3: Compile and Install Using Cygwin

Step 4: Verify Installation on Windows (Cygwin)

Tips and Common Issues

5.3. Linux System BART Installation

Step 1: Environment Preparation

Step 2: Download BART Source Code or Release Package

Step 3: Compile Install, Enable GPU Acceleration (Optional)

Step 4: Install Python Interface (Optional)

Step 5: Verify Installation is Successful

6. Install MONAI

6.1. MONAI Introduction

6.2. Pre-Installation Preparation

6.3. Install PyTorch

6.4. Install MONAI

6.5. Installation Verification

7. Install NiBabel

7.1. NiBabel Introduction

7.2. Install Current Stable Version (Recommended)

7.3. Install Latest Development Version

7.4. Install in "Editable Mode" (Used when Developing NiBabel Source Code)

7.5. Test NiBabel

1. Run Complete Tests During Development (Recommended for Developers Using tox)

2. Test Installed NiBabel