Multi-parametric quantitative in vivo spinal cord MRI with unified signal readout and image denoising

Abstract

Multi-parametric quantitative MRI (qMRI) of the spinal cord is a promising non-invasive tool to probe early microstructural damage in neurological disorders. It is usually performed in vivo by combining acquisitions with multiple signal readouts, which exhibit different thermal noise levels, geometrical distortions and susceptibility to physiological noise. This ultimately hinders joint multi-contrast modelling and makes the geometric correspondence of parametric maps challenging. We propose an approach to overcome these limitations, by implementing state-of-the-art microstructural MRI of the spinal cord with a unified signal readout in vivo (i.e. with matched spatial encoding parameters across a range of imaging contrasts). We base our acquisition on single-shot echo planar imaging with reduced field-of-view, and obtain data from two different vendors (vendor 1: Philips Achieva; vendor 2: Siemens Prisma). Importantly, the unified acquisition allows us to compare signal and noise across contrasts, thus enabling overall quality enhancement via multi-contrast image denoising methods. As a proof-of-concept, here we provide a demonstration with one such method, known as Marchenko-Pastur (MP) Principal Component Analysis (PCA) denoising. MP-PCA is a singular value (SV) decomposition truncation approach that relies on redundant acquisitions, i.e. such that the number of measurements is large compared to the number of components that are maintained in the truncated SV decomposition. Here we used in vivo and synthetic data to test whether a unified readout enables more efficient MP-PCA denoising of less redundant acquisitions, since these can be denoised jointly with more redundant ones. We demonstrate that a unified readout provides robust multi-parametric maps, including diffusion and kurtosis tensors from diffusion MRI, myelin metrics from two-pool magnetisation transfer, and T1 and T2 from relaxometry. Moreover, we show that MP-PCA improves the quality of our multi-contrast acquisitions, since it reduces the coefficient of variation (i.e. variability) by up to 17% for mean kurtosis, 8% for bound pool fraction (myelin-sensitive), and 13% for T1, while enabling more efficient denoising of modalities limited in redundancy (e.g. relaxometry). In conclusion, multi-parametric spinal cord qMRI with unified readout is feasible and provides robust microstructural metrics with matched resolution and distortions, whose quality benefits from multi-contrast denoising methods such as MP-PCA.

Keywords: Marchenko-Pastur PCA denoising; Multi-parametric MRI; Quantitative MRI; Signal readout; Spinal cord.

Fig. 1

Accuracy and precision of different denoising strategies as obtained from percentage relative errors (percentage errors between denoised signals and noise-free ground truth signals) in simulations. Panels A to D (top) show median percentage at different SNR levels, and represent a measure of accuracy (the closer to zero, the higher the accuracy; DW imaging in A, qMT imaging in B, IR imaging in C, mTE imaging in D). Panels E to H (bottom) show percentage relative error interquartile ranges at different SNR levels, and represent a measure of precision (the closer to zero, the higher the precision; DW imaging in E, qMT imaging in F, IR imaging in G, mTE imaging in H). The SNR is evaluated with respected to the white matter signal on the synthetic DW scan at b ​= ​0.

Fig. 2

Top: examples of noise-free (A), noisy (B) and denoised (C) matrices from the synthetic spinal cord phantom. Bottom (D): SV decomposition of the noise-free and noisy matrices shown in A and B, alongside MP-PCA cut off (i.e. edge of noisy SV MP distribution). MP-PCA nullifies all SVs starting from the cut off to the right, while it preserves those to the left. The figure reports results from the simulation conducted with Gaussian noise at an SNR of 15, and considers joint denoising of the whole set of 131 MRI measurements from one spinal cord slice made of 44 voxels (concatenation of DW, qMT, IR and mTE imaging).

Fig. 3

Examples of MP-PCA denoising in one subjects who was scanned with vendor 1. Panels A, B, C, D show raw and denoised images, obtained according to different strategies. DW imaging: panel A; qMT imaging: panel B; IR imaging: panelC; mTE imaging: panelD. Anterior, Posterior, Right, Left respectively indicate subject’s anterior, posterior parts and right and left sides.

Fig. 4

Examples of MP-PCA denoising in two subjects, scanned with vendor 2 respectively in New York and in Montreal. Panels A, B, C, D show raw and denoised images, obtained according to different strategies. DW imaging: images in panels A and C; mTE imaging: images in panels B and D. Ant., Post., Right, Left respectively indicate subject’s anterior, posterior parts and right and left sides.

Fig. 5

Examples of quantitative maps from vendor 1. From top to bottom: FA, MD, MK (DW imaging); BPF, k (qMT imaging); T1 (IR imaging); T2 (mTE imaging). Different rows illustrate the metrics obtained according to different denoising strategies (no denoising; independent denoising of each modality; various combinations of joint multi-modal denoising). Quantitative maps are overlaid onto the mean non-DW image and shown within the cord only. The same anatomical conventions with regard to subject’s anterior, posterior parts and right and left sides as in Fig. 3 are used.

Fig. 6

Examples of quantitative maps from vendor 2 (Siemens Prisma system located in New York, USA). From top to bottom: FA, MD, MK (DW imaging); T2 (mTE). Different rows illustrate the metrics obtained according to different denoising strategies (no denoising; independent denoising of each modality; various combinations of joint multi-modal denoising). Quantitative maps are overlaid onto the mean non-DW image and shown within the cord only. The same anatomical conventions with regard to subject’s anterior, posterior parts and right and left sides as in Fig. 4 are used.

Fig. 7

Examples of quantitative maps from vendor 2 (Siemens Prisma system located in Montreal, Canada). From top to bottom: FA, MD, MK (DWI); T2 (mTE). Different rows illustrate the metrics obtained according to different denoising strategies (no denoising; independent denoising of each modality; various combinations of joint multi-modal denoising). Quantitative maps are overlaid onto the mean non-DW image and shown within the cord only. The same anatomical conventions with regard to subject’s anterior, posterior parts and right and left sides as in Fig. 4 are used.