4D Flow MRI Sequence Implementation

This project implements a 4D phase-contrast MRI sequence for velocity encoding using the pypulseq framework. The sequence incorporates Compressed Sensing for acceleration and Respiratory Controlled Adaptive k-space Reordering (ReCAR) for motion artifact reduction.

Overview

4D flow MRI (time-resolved 3D phase-contrast MRI) is a powerful technique for measuring and visualizing blood flow dynamics in three dimensions over time. This implementation provides a flexible framework for designing and optimizing 4D flow sequences with advanced features:

• Velocity Encoding: Bipolar gradients for encoding velocity in three spatial dimensions
• Compressed Sensing: Variable-density sampling patterns for accelerated acquisition
• ReCAR: Respiratory Controlled Adaptive k-space Reordering for motion artifact reduction
• Visualization: Tools for visualizing the sequence and k-space sampling patterns

Installation

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate 

4d_flow_mri/
│
├── models/                  # Core sequence components and physics models
│   ├── __init__.py
│   ├── sequence_params.py   # Sequence parameters definition
│   ├── gradient_lib.py      # Gradient waveform generation
│   ├── rf_lib.py            # RF pulse generation
│   ├── velocity_encoding.py # Velocity encoding strategies
│   └── compressed_sensing.py # CS sampling patterns
│
├── controllers/             # Sequence control logic
│   ├── __init__.py
│   ├── sequence_builder.py  # Main sequence assembly
│   ├── recar_controller.py  # ReCAR implementation
│   ├── cs_controller.py     # Compressed sensing controller
│   └── export_controller.py # Sequence export functionality
│
├── views/                   # Visualization and UI components
│   ├── __init__.py
│   ├── sequence_plot.py     # Sequence visualization
│   └── k_space_viewer.py    # k-space sampling pattern visualization
│
├── utils/                   # Utility functions
│   ├── __init__.py
│   ├── pulseq_utils.py      # Helpers for pypulseq
│   └── math_utils.py        # Mathematical helper functions
│
├── config/                  # Configuration files
│   ├── __init__.py
│   ├── default_config.py    # Default sequence parameters
│   └── system_config.py     # MRI system specifications
│
├── tests/                   # Unit tests
│   ├── __init__.py
│   ├── test_velocity_encoding.py
│   └── test_compressed_sensing.py
│
├── examples/                # Example implementations
│   ├── basic_4d_flow.py     # Basic 4D flow sequence
│   ├── cs_4d_flow.py        # 4D flow with compressed sensing
│   └── full_4d_flow.py      # Complete implementation with CS and ReCAR
│
├── main.py                  # Main entry point
├── requirements.txt         # Project dependencies
└── README.md                # Project documentation 

This will create a 4D flow sequence with default parameters and export it to output/4d_flow_cs_recar.seq.

Custom Parameters You can customize the sequence parameters by modifying the config/default_config.py file or by passing parameters to the SequenceParams class:

from models.sequence_params import SequenceParams

params = SequenceParams()
params.update(
    fov=[280e-3, 280e-3, 140e-3],
    matrix_size=[192, 128, 32],
    venc=150e-2,  # 150 cm/s
    acceleration_factor=6,
    n_cardiac_phases=20
)  

Examples

The examples/ directory contains several example implementations:

• basic_4d_flow.py: A basic 4D flow sequence without acceleration
• cs_4d_flow.py: A 4D flow sequence with Compressed Sensing
• full_4d_flow.py: A complete 4D flow sequence with CS and ReCAR

Key Features

Velocity Encoding

The sequence uses bipolar gradients for velocity encoding in three spatial dimensions. The velocity encoding value (VENC) can be customized to match the expected flow velocities.

from models.velocity_encoding import create_flow_encoding_gradients

flow_encodings = create_flow_encoding_gradients(
    venc=150e-2,  # 150 cm/s
    system=system,
    flow_directions=[True, True, True]  # Encode in x, y, z directions
)

Compressed Sensing

The sequence uses variable-density sampling patterns for accelerated acquisition. Two sampling patterns are implemented:

Variable-density Poisson disk sampling Phyllotaxis sampling (golden angle-based)

ReCAR (Respiratory Controlled Adaptive k-space Reordering)

ReCAR adaptively reorders k-space acquisition based on respiratory position to reduce motion artifacts. The implementation includes a navigator echo for respiratory motion tracking.

Visualization

The project includes tools for visualizing the sequence and k-space sampling patterns:

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

The pypulseq team for providing the framework for MRI sequence development The authors of the referenced papers for their contributions to 4D flow MRI techniques