Speaker
Description
Introduction: We have recently demonstrated$^1$ a compressed-sensing (CS)-based undersampling method capable of improving signal-to-noise ratio and image quality of low field images. An optimal choice of pulse sequence would reduce undersampling artefacts and improve image quality; in this work, different sampling patterns in k-space for the X-Centric$^2$ and Sectoral$^{3,4}$ sequences are investigated at high acceleration factors (AF = 7, 10, 14).
Method: The X-Centric sequence acquires each half of k-space separately, in the readout direction, reducing signal loss from diffusion and relaxation. Both halves normally acquire the same phase-encode lines in k-space (non-alternating), but they can also sample a unique set of lines (alternating). The Sectoral sequence splits a circular area of k-space into sectors (here, 64), and acquires each sector from the centre-out, oversampling the contrast-rich centre. The proposed sampling pattern consists of stopping each sector prematurely, ensuring the undersampling is confined to the edges of k-space.
In-vitro $^1$H MRI was performed at 73.5mT. Seven sets of 9 images each were acquired with X-Centric: one set per AF for each sampling pattern, and one fully-sampled set to be retrospectively undersampled using the proposed Sectoral sampling. The Fourier-transformed (FT) images were compared to the CS-based reconstructions using the structural similarity index (SSI); all images were 128px$^2$, FOV=8cm$^2$.
Results: The FT images acquired using X-Centric had SSI scores around 35%; however, the FT Sectoral images had a SSI score of 96% and virtually no artefacting, with only slight blurring. The CS reconstructions of all 3 sampling patterns had SSI scores around 87%, with Sectoral exhibiting fewer artefacts.
Conclusion: Although the CS reconstructions of all 3 proposed sampling patterns had similar SSI scores and artefacting, in line with our previous work, the direct FT images of Sectoral were free of artefacts, comparable to the fully-sampled images, even at AF=14 (only 7% of k-space): the artefacts in the CS image are likely due to over-fitting the reconstruction parameters. These results suggest that the proposed Sectoral sampling pattern is well suited for accelerated low field MRI.
References:
1 Perron, S. et al. ISMRM (2022); 2 Ouriadov, A.V. et al. MRM. (2017); 3 Khrapitchev, A. A., et al. JMR (2006); 4 Perron, S. et al. JMR (2022).
| Keyword-1 | MRI |
|---|---|
| Keyword-2 | Pulse Sequence |
| Keyword-3 | Undersampling |
