3D-Dynamic-velocity

A collection of Python tools for processing 4D flow MRI data, including reconstruction from k-space, velocity calculation, and 3D strain analysis.

Requirements

• Python 3.x
• NumPy
• SciPy
• Matplotlib
• tqdm
• BART Toolbox (for reconstruction)

For BART toolbox installation, follow instructions at BART website.

Data Processing Pipeline

The toolkit consists of two main scripts:

1. Reco_GPU.py - Reconstructs images from k-space data and calculates velocity components
2. calc_strain.py - Calculates 3D strain from velocity data

Step 1: Reconstruction and Velocity Calculation

python Reco_GPU.py /path/to/data_dir --venc 20

Arguments

• data_dir: Directory containing .cfl/.hdr k-space data files
• --venc: VENC value in cm/s (default: 20)

This script will:

1. Find all .cfl/.hdr pairs in the data directory
2. Reconstruct images using BART's parallel imaging tools
3. Calculate velocity components (x, y, z) from phase differences
4. Create a mask based on magnitude data
5. Save all results as .npy files in the same directory

Step 2: Strain Calculation

python calc_strain.py --data-path /path/to/data --config config.json

Required Arguments

• --data-path: Directory containing velocity data files
• --config: Path to JSON configuration file with parameters

Optional Arguments

• --prefix: Prefix of velocity data files (default: "DATA")
• --mask: Path to mask file (.npy)
• --roi-mask: Path to region of interest mask file (.npy)
• --output-eig: Output eigenvalue file (.npy) (default: "Eig_v_output.npy")
• --output-disp-x: Output displacement x file (.npy)
• --output-disp-y: Output displacement y file (.npy)
• --output-disp-z: Output displacement z file (.npy)
• --output-plot: Output strain plot file (.png)
• --slice-index: Slice index for 2D analysis
• --no-display: Do not display plots

Configuration File

Example config.json:

{
  "CardiacNumberOfImages": 27,
  "InPlaneResolution": [1.5, 1.5],
  "SliceThickness": 1.5,
  "RepetitionTime": 6.7,
  "VelocityDirectionMatrix": [1, 1, 1]
  "AffineMatrix":np.eye(4)
}

The VelocityDirectionMatrix is a diagonal matrix represented as three values that controls the sign of each velocity component (x, y, z) obtained from the gradient probing sequence. Use 1 for positive, -1 for negative direction.

The affine matrix is extracted from the raw data after conversion to ISMRMRD data with the get_matrix script, if none is provided a identy matrix is used as stand in.

Example Workflow

1. Data Reconstruction:

   python Reco_GPU.py  /path/to/data --venc 25

2. Strain Calculation:

   python calc_strain.py --data-path  /path/to/data --config settings.json --output-plot strain_results.png

Output

• Velocity components (.npy files)
• Displacement vectors (.npy files, optional)
• Strain eigenvalues (.npy file)
• Strain plot (.png file, optional)
• Nifti of the reconstructed data and velocity maps (.nii file, optional)