Figure 2 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Image reconstruction, processing, and analysis pipelines. (a) The data were put through a series of steps to extract and reorder the k-space from the scanner and corrected for thermal drift and denoising in k-space. (b) The post-processing involved off-resonance correction and low-pass filtering to yield the images for analysis. (c) Different slices were chosen to test for signal-to-noise ratio (SNR), image uniformity, and geometrical distortion and were manually segmented to provide input for computing the relevant image quality metric.
Source publication
We investigated the repeatability of image quality metrics such as SNR, image uniformity, and geometrical distortion at 0.05T over ten days and three sessions per day. The measurements included temperature, humidity, transmit frequency, off-resonance maps, and 3D turbo spin echo (TSE) images of an in vitro phantom. This resulted in a protocol with...
Contexts in source publication
Context 1
... parameters were selected to assess the repeatability of the scans across different imaging sessions over ten days and involved measurements on different slices. Supplementary Figure 2(a-c) displays the representative specific slices, without offresonance correction, chosen to evaluate SNR, image uniformity, and geometric accuracy in this study. The Pro-MRI phantom's construction is similar to the American College of Radiology phantom, a standard phantom used in MRI. ...
Context 2
... SNR was computed as the ratio of the mean signal intensity of the phantom ROI to the standard deviation of the background noise. To evaluate IU, we chose slice 16 and computed the standard deviation with the phantom region of interest (ROI, red ROI in Figure 2c; the lesser the standard variation, the more uniform the intensity). For GA, we chose slice number 8 and calculated the eccentricity of the phantom ROI, which measures the degree of deviation from a perfect circle. ...
Context 3
... reconstruction: Figure 2(a-c) shows the reconstruction, post-processing, and image analysis pipelines. The k-space data underwent preprocessing steps, including drift correction and k-space filtering (squared sine-bell). ...
Context 4
... Figure 2. Raw data (not corrected for off-resonance) slices considered for image quality metrics: a) signal-to-noise ratio -slice 10; b) image non-uniformity quantified by the standard deviation inside the inner circle (ideal = 0)-slice 16; c) geometric accuracy quantified by eccentricity (ideal = 0) -slice 8 ...
Similar publications
Examining and comparing the image quality of degenerative cervical spine diseases through the application of three MRI sequences; the Two-Dimension T2 Weighed Turbo Spin Echo (2D T2W TSE), the Three-Dimension T2 Weighted Turbo Spin Echo (3D T2W TSE), and the T2 Turbo Field Echo (T2_TFE). Thirty-three patients who were diagnosed as having degenerati...
Objective
The objective of this study is to prospectively compare quantitative and subjective image quality, scanning time, and diagnostic confidence between a new deep learning-based reconstruction(DLR) algorithm and standard MRI protocol of lumbar spine.
Materials and methods
Eighty healthy volunteers underwent 1.5T MRI examination of lumbar spi...
Citations
... The phantom dataset comprises T1-weighted axial Pro-MRI phantom images acquired using a single-coil 0.05T Multiwave MGNTQ MRI scanner during a repeatability study (Aggarwal P. P. K. et al., 2023). The images used were not corrected for geometric distortion. ...
Low-field MRI is gaining interest, especially in low-resource settings, due to its low cost, portability, small footprint, and low power consumption. However, it suffers from significant noise, limiting its clinical utility. This study introduces native noise denoising (NND), which leverages the inherent noise characteristics of the acquired low-field data. By obtaining the noise characteristics from corner patches of low-field images, we iteratively added similar noise to high-field images to create a paired noisy-clean dataset. A U-Net based denoising autoencoder was trained on this dataset and evaluated on three low-field datasets: the M4Raw dataset (0.3T), in vivo brain MRI (0.05T), and phantom images (0.05T). The NND approach demonstrated improvements in signal-to-noise ratio (SNR) of 32.76%, 19.02%, and 8.16% across the M4Raw, in vivo and phantom datasets, respectively. Qualitative assessments, including difference maps, line intensity plots, and effective receptive fields, suggested that NND preserves structural details and edges compared to random noise denoising (RND), indicating potential enhancements in visual quality. This substantial improvement in low-field imaging quality addresses the fundamental challenge of diagnostic confidence in resource-constrained settings. By mitigating the primary technical limitation of these systems, our approach expands the clinical utility of low-field MRI scanners, potentially facilitating broader access to diagnostic imaging across resource-limited healthcare environments globally.
