Target Audience

Clinicians and scientists interested in understanding the benefits of non-Cartesian sampling, common applications, and important considerations for selecting a scheme and mitigating challenges.

Objectives

This talk will provide information about the benefits and drawbacks of choosing a non-Cartesian sampling method over Cartesian sampling. Key points include:

• Advantages of non-Cartesian sampling
• Disadvantages of non-Cartesian sampling and methods for mitigating those issues
• Common non-Cartesian sampling patterns and clinical applications

Advantages and Pitfalls

Scan Efficiency: Conventional Cartesian sampling, where one line of k-space is acquired per TR, requires a relatively long time to fully sample k-space. Trajectories that curve, such as spirals [1,2], can cover the desired portion of k-space in fewer excitations, reducing the total acquisition time. However, non-Cartesian sampling can lead to reduction in SNR due to non-uniform noise weighting.

Undersampling Artifacts: Undersampled Cartesian k-space data results in replicates of the object in image space, which can obscure important features of the image. Undersampling artifacts for non-Cartesian sampling patterns manifest in different ways, which may be advantageous for image interpretation. Furthermore, these artifacts tend to be less coherent with the desired signal, enabling more advanced imaging methods such as MR fingerprinting or compressed sensing. In Cartesian sampling, an anti-aliasing filter may be applied to the readout direction to reduce artifacts. However, due to the lack of a single readout direction, this is not an option in non-Cartesian methods.

Motion: Motion during Cartesian sampling results in ghosting artifacts along the phase encode direction. Non-Cartesian sampling methods can result in reduced or less coherent motion artifacts. Methods that repeatedly sample the center of k-space provide the opportunity to track motion for use in motion compensation or correction methods. Non-Cartesian sampling methods also tend to have lower gradient moments, which result in less flow-related signal loss.

Main Field Inhomogeneity: For Cartesian sampling, main field inhomogeneities can cause image distortions. Some non-Cartesian schemes are more robust to off-resonance effects while others yield reduced image quality. For example, rosette trajectories [3] are robust to off-resonance effects because they incoherently distribute off-resonant signal across the image, but spirals suffer from image blurring. Many correction methods exist for minimizing the effects of off-resonance [4-6].

Reconstruction: Image reconstruction for Cartesian data is straightforward because the FFT can be applied directly to the k-space data. Non-Cartesian data is more complex to reconstruct because the FFT cannot be directly applied, yielding higher computational demand. Common methods for reconstructing non-Cartesian k-space data include gridding [7] or the non-uniform FFT (NUFFT) [8]. Other methods include back-projection and compressed sensing.

Hardware Imperfections: Non-Cartesian sampling patterns are more susceptible to gradient delays and Eddy currents. However, there are methods for measuring and correcting for these non-idealities [9,10]. These correction methods add overhead to implementing non-Cartesian MRI but are usually outweighed by the benefits.

Common Methods and Applications

Common 2D non-Cartesian sampling methods include radial, spiral, and PROPELLER [11]. These methods can be extended to 3D with trajectories such as stack of stars, stack of spirals, and VIPR [12]. There are many more 2D and 3D trajectories beyond this list, some of which will be covered in this talk.

Nearly all MRI applications benefit from the reduced scan time offered by non-Cartesian sampling. There are also applications that make use of more specific advantages of non-Cartesian sampling. For instance, cardiac and abdominal MRI benefits from motion tolerance as well as motion tracking features. This talk will cover a handful of clinical applications and the associated non-Cartesian sampling methods.

Acknowledgements

No acknowledgement found.