Despite a half-century of advancements, global magnetic resonance imaging (MRI) accessibility remains limited and uneven, hindering its full potential in health care. Initially, MRI development focused on low fields around 0.05 Tesla, but progress halted after the introduction 1.5 Tesla whole-body superconducting scanner 1983. Using permanent magnet deep learning for electromagnetic interference elimination, we developed that operates using standard wall power outlet without radiofrequency...

Purpose Recent development of ultra‐low‐field (ULF) MRI presents opportunities for low‐power, shielding‐free, and portable clinical applications at a fraction the cost. However, its performance remains limited by poor image quality. Here, computational approach is formulated to advance ULF MR brain imaging through deep learning large‐scale publicly available 3T data. Methods A dual‐acquisition 3D superresolution model developed 0.055 T. It consists cross‐scale feature extraction, attentional...

In recent years, there has been an intensive development of portable ultralow-field magnetic resonance imaging (MRI) for low-cost, shielding-free, and point-of-care applications. However, its quality is poor scan time long. We propose a fast acquisition deep learning reconstruction framework to accelerate brain MRI at 0.055 tesla. The consists single average three-dimensional (3D) encoding with 2D partial Fourier sampling, reducing the T1- T2-weighted protocols 2.5 3.2 minutes, respectively....

We aim to explore the feasibility of head and neck time-of-flight (TOF) magnetic resonance angiography (MRA) at ultra-low-field (ULF). TOF MRA was conducted on a highly simplified 0.05 T MRI scanner with no radiofrequency (RF) shielding. A flow-compensated three-dimensional (3D) gradient echo (GRE) sequence tilt-optimized nonsaturated excitation RF pulse, multislice two-dimensional (2D) GRE sequence, were implemented for cerebral artery vein imaging, respectively. For carotid jugular 2D...

Purpose To develop a joint denoising method that effectively exploits natural information redundancy in MR DWIs via low‐rank patch matrix approximation. Methods A is introduced to jointly reduce noise DWI dataset by exploiting nonlocal self‐similarity as well local anatomical/structural similarity within multiple 2D acquired with the same anatomical geometry but different diffusion directions. Specifically, for each small 3D reference sliding DWI, similar patches are searched matching image...

Abstract Purpose To demonstrate magnetization transfer (MT) effects with low specific absorption rate (SAR) on ultra‐low‐field (ULF) MRI. Methods MT imaging was implemented by using sinc‐modulated RF pulse train (SPT) modules to provide bilateral off‐resonance irradiation. They were incorporated into 3D gradient echo (GRE) and fast spin (FSE) protocols a shielding‐free 0.055T head scanner. first verified phantoms. Brain conducted in both healthy subjects patients. Results clearly observed...

Purpose To develop a robust parallel imaging reconstruction method using spatial nulling maps (SNMs). Methods Parallel null operations (PRUNO) is k‐space where system derived null‐subspace bases of the calibration matrix. ESPIRiT extends PRUNO subspace concept by exploiting linear relationship between signal‐subspace and coil sensitivity characteristics, yielding hybrid‐domain approach. Yet it requires empirical eigenvalue thresholding to mask information sensitive signal‐ division. In this...

Predicting brain age from structural MRI (sMRI) is potentially valuable as the deviation of predicted chronological can be a biomarker for characterising health conditions. Currently, extensive pre-processing sMRI data required most deep learning methods. This study presents multi-task contrastive framework simultaneous prediction and gender classification minimally processed, noisy 3D T1-weighted images. By including task supervised learning, we demonstrate that leveraging information in...

3D MRI data contains more redundant information than 2D data, which is favourable for reconstruction. However, deep learning reconstruction of remains to be explored due the computational burden that scales exponentially with spatial dimensions. This study presents a method reconstruct single-channel uniform undersampling along two phase-encoding directions, in conventional multi-channel parallel imaging methods are generally not applicable. The results demonstrate robust at high...

Motivation: Ultra-low-field (ULF) MRI technology holds significant promise for advancing medical imaging by offering low-cost and portable solutions point-of-care applications. These advancements have the potential to improve access in resource-limited settings, thereby benefiting underserved populations enhancing diagnostic capabilities ultimately patient care. Goal(s): The implementation of a highly efficient protocol ULF MRI. Approach: A 3D bSSFP was implemented optimized. Results: study...

Motivation: High, isotropic resolution (e.g., 1mm) is desirable for lesion detection and biomarkers extraction cognitive disorders. However, ultra-low-field (ULF) MRI severely suffers from low spatial signal-to-noise ratio. Goal(s): To investigate the potential of 3D deep learning in generating <=1mm results 2D partial Fourier-sampled, low-resolution noisy brain images acquired our custom-made 0.055T scanner. Approach: We advanced Fourier reconstruction super-resolution method (PF-SR)...

Motivation: The recent resurgence of ultra-low-field MRI (i.e., below 0.1 T) is showing great promise for future clinical applications due to its low cost, portability, and accessibility, potentially advancing neck MRA evaluating diagnosing carotid diseases in point-of-care scenarios low/mid-income countries. Goal(s): To explore the using TOF technique at 0.05 Tesla. Approach: Flow-compensated 2D GRE sequences with without flow saturation multi-slice scans. Results: Carotid...

Motivation: Deep learning (DL) is a powerful tool for MR image formation tasks and data at ultra-low-field (ULF) strength has significantly lower SNR than high-field. Goal(s): Enhancing the quality of ULF knee c-spine 0.05T via DL reconstruction. Approach: We extend our recently developed 3D partial Fourier reconstruction superresolution (PF-SR) method on PF-sampled low-resolution noisy brain to data. Results: The preliminary results demonstrate PF-SR, trained synthetic simulated from...

Motivation: While the emerging ULF MRI shows potential of low-cost and point-of-care imaging applications, its image quality is poor scan time long. Goal(s): To reduce brain through deep learning reconstruction from partial Fourier uniformly undersampled data. Approach: We proposed a DL method for fast 3D at 0.055T by applying to k-space data, achieving speed up 2x over our newly developed superresolution (PF-SR) method. Results: Our preliminary results show could noise, artifacts, enhance...

Motivation: To develop low-cost and patient-friendly MRI scanners to address global healthcare disparities. Goal(s): demonstrate cervical spine (C-spine) on a RF shielding-free 0.05T scanner. Approach: Typical imaging protocols were implemented newly developed The scanner is compact, shielding-free, acoustically quiet during scanning. Further, deep learning electromagnetic interference (EMI) elimination method data-driven reconstruction strategy designed. Results: EMI effectively removed...

We present a U-Net based deep learning model to estimate the multi-channel ESPIRiT maps directly from uniformly-undersampled multi-slice MR data. The is trained with hybrid loss function using fully-sampled axial brain datasets same receiving coil system. proposed robustly predicted k-space and cardiac data, yielding highly comparable performance reconstruction acquired reference maps. Our method presents general strategy for calibrationless parallel imaging through protocol specific

Recent development of ultra-low-field (ULF) MRI presents opportunities for low-cost and portable imaging in point-of-care scenarios or/and low- mid-income countries. Magnetic resonance elastography (MRE) is an essential part MR abdominal especially chronic liver diseases. In this study, we explore the MRE at 0.055 Tesla. We demonstrate feasibility based on phantom experiments

Recently, there has been an impetus to develop ultra-low-field (ULF) MRI technologies, which present opportunities for low-cost and portable imaging in point-of-care scenarios. Balanced steady-state free precession (bSSFP) is a time-efficient sequence yet its feasibility at ULF remains unexplored. In this study, we implemented bSSFP 0.055T, successfully demonstrated phantom, brain extremity imaging.

We present a CNN-based deep learning model to reconstruct the cardiac cine images from undersampled single-channel 4D MR data. The wavelet transform and spatial-temporal attention mechanisms are introduced in model. proposed could recover wall motion more robustly than low-rank plus sparsity (i.e., L+S) reconstruction method. This approach presents one promising solution for accelerated dynamic imaging with single channel through spatial temporal domains.

Recent development of ultra-low-field (ULF) MRI presents opportunities for low-cost and portable brain imaging in point-of-care scenarios or/and low- mid-income countries. Magnetic resonance angiography (MRA) is an essential part MR neuroimaging protocols especially stroke assessment, yet its feasibility at ULF remains unknown. In this study, we explore the time-of-flight MRA 0.055 Tesla. We demonstrate cerebral using flow-compensated gradient echo sequences, enabling visualization main...

At high field, magnetization transfer (MT) imaging suffers from the specific absorption ratio (SAR) issue due to usage of power off-resonance MT RF pulses, and on-resonance saturation caused by B0 field inhomogeneity. ultra-low-field (ULF), low SAR absolute inhomogeneity (in Hz) greatly facilitates application strong versatile pulses without confounding in practice. We demonstrate brain at ULF first time using a 0.055 Tesla MRI platform with an extremely SAR.

We develop a novel parallel imaging reconstruction method by extracting null-subspace bases of calibration data/matrix to calculate image-domain spatial nulling maps that contain both coil sensitivity and finite image support information. Images are reconstructed solving system formed multi-channel without any masking-related procedure (i.e., in existing SENSE/ESPIRiT for minimizing noise propagation). demonstrate this with 2D brain, knee cardiac data under various conditions, yielding...

MRI scans are commonly performed inside a fully-enclosed RF shielding room, posing stringent installation requirement and unnecessary patient discomfort. This study develops strategy of active EMI sensing deep learning MR signal prediction using residual U-Net for shielding-free MRI. We implemented it on an ultra-low-field 0.055T head scanner. Our experimental results demonstrated that this could directly accurately predict EMI-free signals from the acquired by receive coil coils. It worked...

Purpose To develop a truly calibrationless reconstruction method that derives An Eigenvalue Approach to Autocalibrating Parallel MRI (ESPIRiT) maps from uniformly‐undersampled multi‐channel MR data by deep learning. Methods ESPIRiT, one commonly used parallel imaging technique, forms the images undersampled k‐space using ESPIRiT effectively represents coil sensitivity information. Accurate map estimation requires quality calibration or autocalibration data. We present U‐Net based learning...

Wave encoding offers high acceleration in parallel imaging by exploring the coil sensitivity variations readout dimension. However, typically preset wave gradients (i.e., sinusoidal trajectory) have never been optimized for specific receiver array coil, thus intrinsically limiting maximum factor. We propose to optimize gradient trajectory a manner minimizing squared L2-norm of off-diagonal elements correlation matrix. To guarantee an allowed slew rate, bandlimited constraint is also...