Super resolution using sparse sampling at portable ultra-low field MR

Abstract

Ultra-low field (ULF) magnetic resonance imaging (MRI) holds the potential to make MRI more accessible, given its cost-effectiveness, reduced power requirements, and portability. However, signal-to-noise ratio (SNR) drops with field strength, necessitating imaging with lower resolution and longer scan times. This study introduces a novel Fourier-based Super Resolution (FouSR) approach, designed to enhance the resolution of ULF MRI images with minimal increase in total scan time. FouSR combines spatial frequencies from two orthogonal ULF images of anisotropic resolution to create an isotropic T2-weighted fluid-attenuated inversion recovery (FLAIR) image. We hypothesized that FouSR could effectively recover information from under-sampled slice directions, thereby improving the delineation of multiple sclerosis (MS) lesions and other significant anatomical features. Importantly, the FouSR algorithm can be implemented on the scanner with changes to the k-space trajectory. Paired ULF (Hyperfine SWOOP, 0.064 tesla) and high field (Siemens, Skyra, 3 Tesla) FLAIR scans were collected on the same day from a phantom and a cohort of 10 participants with MS or suspected MS (6 female; mean ± SD age: 44.1 ± 4.1). ULF scans were acquired along both coronal and axial planes, featuring an in-plane resolution of 1.7 mm × 1.7 mm with a slice thickness of 5 mm. FouSR was evaluated against registered ULF coronal and axial scans, their average (ULF average) and a gold standard SR (ANTs SR). FouSR exhibited higher SNR (47.96 $\pm$ 12.6) compared to ULF coronal (36.7 $\pm$ 12.2) and higher lesion conspicuity (0.12 $\pm$ 0.06) compared to ULF axial (0.13 $\pm$ 0.07) but did not exhibit any significant differences contrast-to-noise-ratio (CNR) compared to other methods in patient scans. However, FouSR demonstrated superior image sharpness (0.025 $\pm$ 0.0040) compared to all other techniques (ULF coronal 0.021 $\pm$ 0.0037, q = 5.9, p-adj. = 0.011; ULF axial 0.018 $\pm$ 0.0026, q = 11.1, p-adj. = 0.0001; ULF average 0.019 $\pm$ 0.0034, q = 24.2, p-adj. < 0.0001) and higher lesion sharpness (-0.97 $\pm$ 0.31) when compared to the ULF average (-1.02 $\pm$ 0.37, t(543) = -10.174, p = <0.0001). Average blinded qualitative assessment by three experienced MS neurologists showed no significant difference in WML and sulci or gyri visualization between FouSR and other methods. FouSR can, in principle, be implemented on the scanner to produce clinically useful FLAIR images at higher resolution on the fly, providing a valuable tool for visualizing lesions and other anatomical structures in MS.

Keywords: Fourier-transform; multiple sclerosis; reconstruction; super-resolution; ultra-low field.

Figure 1

Examples of how sampling frequency influences the quality of MRI images. A coronal slice of a standard hyperfine phantom, scanned at 3T and downsampled to 1.6x1.6x1.6mm voxel sizes, was Fast Fourier transformed (FFT) to obtain frequency space data. The inverse FFT (IFFT) of (A) only the central portion of the frequency space produces a blurred image that is rich in contrast as seen in the phantom; (B) the periphery of the frequency space produces images with clear edges of the phantom but lacking in contrast between fluid and background; and (C) the full frequency space produces the highest quality images.

Figure 2

Overview of workflow to create FouSR images. (A) Same-day FLAIR scans were acquired at ULF (64 mT) and HF (3 T). ULF FLAIR images were acquired in 2 planes, coronal and axial, but are shown here in triplanar reformation. (B) All FLAIR images were interpolated to 1.7 mm isotropic. ULF images were brain extracted using SynthStrip (9). HF images were cropped to remove neck and BRAIN extracted using FSL’s tools (RobustFOV, BET2) (10). (C) ULF FLAIR scans were nonlinearly registered to the down-sampled HF FLAIR using the ANTs multivariate template construction tool (11). The transformation matrix was applied to the non-brain extracted but up-sampled ULF images using nearest neighbor interpolation. (D) The ULF images were Fast Fourier transformed and the missing high-frequency components in the under-sampled direction of the ULF axial FLAIR was replaced with that from the ULF coronal FLAIR. The new frequencies were inversely Fast Fourier transformed to image space.

Figure 3

In a standard Hyperfine phantom, FouSR shows improved edge sharpness and image quality compared to ULF FLAIR scans. Representative (A) axial, (B) coronal, and (C) sagittal slices (either directly acquired or reformatted) are shown for ULF coronal FLAIR, ULF axial FLAIR, FouSR, ULF average, and 3 T FLAIR (left to right). Red arrows highlight an example area across each plane where FouSR demonstrates improved resolution and image sharpness compared to other methods.

Figure 4

FouSR improves image quality across all three planes. Qualitative visualization of (A) axial, (B) coronal, and (C) sagittal slices in an MS case with high lesion burden in ULF FLAIR images, the average of ULF FLAIR images, after FouSR algorithm and after ANTs SR algorithm is shown. Arrows point to artifacts from partial voluming along the slice direction that is seen in the coronal and axial ULF (orange arrows) but reduced in FouSR and average images (yellow arrows) and absent in ANTs-SR images (green arrows).

Figure 5

FouSR improves image quality across all three planes. Qualitative visualization of (A) axial, (B) coronal, and (C) sagittal slices in an MS case with high lesion burden in ULF FLAIR images, the average of ULF FLAIR images, after FouSR algorithm and after ANTs SR algorithm is shown. Arrows point to artifacts from partial voluming along the slice direction that is seen in the coronal and axial ULF (orange arrows) but reduced in FouSR and average images (yellow arrows) and absent in ANTs-SR images (green arrows).

Figure 6

In-vivo quantitative assessments of (A) SNR, (B) CNR, (C) image sharpness, and (D) lesion sharpness between individual images, average image, FouSR and ANTs-SR in 10 adults with MS show that FouSR produces the sharpest images. (^*p < 0.05 in pairwise mean difference).

Figure 7

Qualitative ratings for white matter lesions (WML) are highest in ULF average, followed by FouSR and ANTs SR. ANTs SR has higher qualitative ratings of sulci and gyri than ULF coronal. For each method, per participant qualitative rating of (A) WML (B) sulci and gyri. (^*p < 0.05, pairwise mean difference).