Keywords: Motion Correction, Motion Correction

Motivation: Motion artifacts are a common source of artifacts in clinical brain imaging.

Goal(s): To facilitate efficient retrospective motion correction for susceptibility-weighted-imaging (SWI).

Approach: A 2 sec motion-free pre-scan and the repeated acquisition of additional k-space encoding lines (guidance lines) were implemented into an GRE-based SWI. Guidance lines are played as an additional gradient-echo before the imaging echo which retains the original scan efficiency. Scout and guidance lines are then used for very rapid, fully separable motion trajectory estimation and correction.

Results: In vivo, reduced motion artifacts and increased image sharpness is demonstrated across several scans with instructed subject motion.

Impact: Scout and guidance-lines based retrospective motion correction is introduced for SWI. Our approach enables efficient motion artifact mitigation while being minimally disruptive to standard clinical protocols and should improve the robustness and reproducibility of clinical brain imaging.

Keywords: Motion Correction, Motion Correction

Motivation: Motion frequently impairs T2 weighted prostate imaging; in the multi-pass acquisitions that are common this results in an objectionable stair-step artifact, and can lead to additional scans and increased exam time.

Goal(s): Reduce the impact of motion occurring in a subset of the multi-pass acquisition.

Approach: Automatic detection of motion (through a mutual-information metric) coupled with replacement of corrupted slices via averaging of neighbors. In the described case, adjacent slices are intentionally overlapped, leading to reasonable results from this approach.

Results: Improvements are clearly visible in both motion-controlled phantom and a subset of volunteer examinations.

Impact: The ability to automatically repair a portion of motion-corrupted exams may enable reduction of supplemental / repeat scans, improving overall exam times. The described technique is not computationally complex, and could be performed inline after scan completion.

Keywords: Parallel Imaging, High-Field MRI

Motivation: To optimize an EDGE-MP2RAGE acquisition at 7T for epilepsy studies.

Goal(s): To determine if it is beneficial to use Universal Pules within EDGE-MP2RAGE at 7T and to find the optimum TI1 to produce the best visulation of the gray-white matter boundary.

Approach: Collect EDGE-MP2RAGE data with conventional RF and Universal Pulses and analyze the gray and white matter signal across the brain.

Results: Overall, it does seem to be beneficial to use Univeral Pulses within EDGE-MP2RAGE, although it does not produce more consistent gray and white matter signal in all brain regions.

Impact: Utilizing Universal Pulses within EDGE-MP2RAGE with the determined optimum TI1=760ms at 7T can improve continuity of the gray-white matter boundary compared to conventional RF pulses, which may allow for detection of focal cortical dysplasia in epilepsy patients.

Keywords: Artifacts, Artifacts

Motivation: Effect of free diffusion introduced image blur is well known but measurable samples always have solid boundaries which results in non-blur like artefacts in ultra-high-resolution images.

Goal(s): Demonstration of diffusion introduced artefact behavior near impermeable barriers.

Approach: A Monte Carlo simulation of particles undergoing a random walk was performed and CTI specific point spread functions (PSFs) and 1D images determined.

Results: It was demonstrated that the PSF yields asymmetric blur and signal enhancement in direct vicinity of the barrier. Simulations clearly indicating the PSF to be dependent on spatial resolution and gradient strength.

Impact: These results will help to prohibit misinterpretation of MR images of small structures with ultra-high-resolution MRI.

Keywords: Motion Correction, Motion Correction, MOLLI、registration

Motivation: Accurate cardiac T1 mapping is crucial for diagnosing heart conditions, yet patient motion can cause misaligned images. We aimed to address this with an automatic registration system.

Goal(s): Develop and validate a high-precision automatic registration system for aligning MOLLI cardiac images.

Approach: We created a system that integrates a GAN-generated virtual MOLLI target (VMT) and a deep-learning-based multi-modal registration method (DL) and applied it to a dataset, using the fitting quality index (FQI) for assessment.

Results: Our findings indicate that while all three tested registration methods improved alignment. Our VMT+DL system consistently performed well in datasets with significant motion, while traditional methods faltered.

Impact: The VMT+DL system offers a robust alternative for cardiac T1 mapping in clinical settings, where patient movement can compromise image registration. It ensures the reliability of diagnostic imaging, which is crucial for patient care in cardiology.

Keywords: Quantitative Imaging, Quantitative Susceptibility mapping

Motivation: This work aims to automate analysis of the newly introduced Iron Microstructure Coefficient (IMC) to facilitate understanding of iron cellular distribution in neurological diseases for large cohort studies.

Goal(s): The goal is to develop and test an automated and widely applicable pipeline for IMC analysis using quantitative susceptibility and R₂^* maps obtained from the same multi-echo GRE sequence.

Approach: The pipeline inputs magnetic susceptibility and R₂^* maps, T1-w data, templates, and regions of interest (ROIs). The output is the IMC value per ROI per subject.

Results: The pipeline successfully executed in 50 minutes without segmentation failure for a cohort of 28 test subjects.

Impact: The automated pipeline accelerates data processing for IMC, providing enhanced standardization in a robust, reproducible, and user-friendly manner. It facilitates large-scale research, driving significant advancements in our understanding of neurological diseases, with the goal of improving accurate diagnosis for patients.

Keywords: Quantitative Imaging, Quantitative Imaging

Motivation: The MULTIPLEX technique could quantify the T1/T2*/PD/Susceptibility maps in a single 3D scan but leads to a long scan time.

Goal(s): This study aimed to develop sparsity regularized Wave-SNMs reconstruction methods to address the issue of slow 3D scanning of the MULTIPLEX technique.

Approach: This work is based on a SNMs reconstruction method with the addition of a sparsity regular term of L1, which utilizes nulling maps with a short calibration time to achieve a 12-fold accelerated imaging.

Results: Sparsity regularized Wave-SNMs reconstruction with Multi-Dimensional Integration quantification accelerate MULTIPLEX by 12-fold into a single scan of 2 minutes.

Impact:  The L1 regularized Wave-SNMs reconstruction could benefits the Mutli-Dimensional lntegration (MDI) quantification to achieve comparable accuracy and robustness as the reference scan with 70% reduction of scan time.

Keywords: Quantitative Imaging, Validation, Reproducibility, model-based reconstruction, sensitivity analysis, state-transition matrix, nonlinear inversion, Bloch equations, quantitative MRI

Motivation: Inverse problems in MRI require estimation of the Bloch equation partial derivatives, however robust computation is challenging. Sensitivity analysis offers accurate and numerically stable derivatives to overcome this.

Goal(s): To replicate and validate previous work, and examine functionality of the BART toolbox and use of collaborative platforms such as GitHub for reproducing research.

Approach: A direct replication was attempted following methodology from a previous ISMRM abstract, and an accompanying preprint and GitHub repository.

Results: Both replicators successfully recreated all abstract figures. Replication of difference quotient and sensitivity analysis derivatives was achieved within the predefined normalised root mean square error tolerances.

Impact: Successful replication further validates novel work in computing partial derivatives of the Bloch Equations. Collaborative platforms such as GitHub can improve existing software and resources when reproducing research. This enables wider dissemination, enhancing ease-of-use for other researchers in future applications.

Keywords: Artifacts, Low-Field MRI

Motivation: Spectrally-encoded multi-spectral imaging (SEMSI) is an approach for distortion-free MRI around metallic implants, but requires multiple-readouts per TR to be clinically feasible.

Goal(s): To determine the benefits of SEMSI over SEMAC at 0.55T, using parameters that are consistent with a future multi-readout SEMSI implementation.

Approach: We used single-readout SEMSI to prospectively mimic the performance of multi-readout SEMSI with high slew rate and readout bandwidth.

Results: We observe that SEMSI simultaneously achieves the expected SNR improvement and artifacts reduction compared to SEMAC.

Impact: SNR improvement is important at 0.55T to examine the tissues near the metallic implants. This prospective study confirms the improvement performance of multi-readout SEMSI which will further improve SNR and scan efficiency.

Keywords: Quantitative Imaging, Precision & Accuracy, Phantoms, PDFF, T1, Multi-center, Multi-vendor

Motivation: Chemical-shift-encoded (CSE)-based proton-density fat-fraction (PDFF) is a highly validated biomarker of liver fat. T1 mapping has been proposed as a biomarker of hepatic fibrosis. However, the reproducibility of PDFF and T1 in the setting of concomitantly varying fat and T1 is poorly understood.

Goal(s): To validate the reproducibility of CSE-based PDFF and MOLLI-based T1 mapping with concomitantly varying fat and T1.

Approach: We conducted a four-center, four-vendor validation study using a quantitative PDFF-T1 phantom.

Results: CSE-based PDFF had good reproducibility, although with increased bias and variability at long T1 values. The reproducibility of MOLLI-based T1 was affected substantially by the presence of fat.

Impact: CSE-based PDFF demonstrated good reproducibility across four centers/vendors, at both 1.5T and 3T. Increased T1 and increased PDFF led to reduced MOLLI-based T1 reproducibility. This multi-center multi-vendor PDFF-T1 phantom validation approach may enable evaluation of improved quantitative MRI methods.

Keywords: Pulse Sequence Design, Atherosclerosis, 3D DWI

Motivation: Non-Cartesian sampling offers advantages in motion robustness and speed, crucial in 3D DWI. 

Goal(s): We aimed to improve 3D DWI using RAZER trajectory to improve the 3D DWI imaging with reducing scan time while preserving imaging quality. 

Approach: We implemented 3D DW RAZER sequence with integrated radial and EPI sampling and conducted 3D DWI studies in a phantom and human carotid arteries.

Results: The RAZER DWI sequence offered a clearer depiction of the carotid arterial structure and reliable ADC measurements without geometric distortion when compared to 2D ss-DW EPI, all within a scan time of less than two minutes.

Impact: This study introduces a motion-robust 3D DWI with faster imaging capabilities, potentially improving vulnerable plaque identification. It addresses conventional DWI limitations, providing a valuable resource for the detection of possible events and improved patient outcomes in ischemic stroke.
 

Keywords: Pulse Sequence Design, Brain

Motivation: MRI produces considerable acoustic noise. PETRA sequence can solve this problem but has limited contrast.

Goal(s): This study improves PETRA sequence by magnetization preparation to achieve widely used T1-weighted imaging with fat suppression and sound reduction. 

Approach: Double asymmetric adiabatic RF pulses are designed to invert aqueous tissues for T1-weighted contrast creation and invert lipidic tissues for fat suppression and combined with PETRA gradient trajectories. 

Results: The brain and knee experiments are conducted to verify the feasibility of the proposed sequence. The fat is decreased by above 70%. The gray-to-white matter contrast and knee cartilage visualization are enhanced compared with non-prepared PETRA sequence.

Impact: The significantly reduced SPL of 67.4 dBA using limited gradient switching, providing better patient scanning comfort, can promote its acceptance to pediatric imaging. The double RF preparation design improves the limited contrast of PETRA.

Keywords: Image Reconstruction, Sparse & Low-Rank Models

Motivation: Gu et al. [3] showed one can obtain comparable performance as the physics-guided deep learning (PG-DL) networks [4] for CS-MRI reconstruction by using multiple wavelets as the regularizers. 

Goal(s): Develop an efficient numerical algorithm for CS-MRI reconstruction with multiple wavelets.

Approach: Study a fast double stochastic proximal method (FDSPM) for compressed sensing MRI (CS-MRI) reconstruction.

Results: Our experiments demonstrate that FDSPM converges in less CPU time than classical CS algorithms for image reconstruction.

Impact: Exploring efficient algorithms for multiple regularizers CS-MRI reconstruction can motivate new efficient network structures that are easy to train.

Keywords: Machine Learning/Artificial Intelligence, Visualization, Adaptive convolution, visual pathway segmentation, deep learning

Motivation: Accurate visual pathway (VP) segmentation is critical for clinical diagnosis and surgical planning. Current deep learning-based methods struggle to capture significant context information, impacting the segmentation precision.

Goal(s): Improve multi-parametric MRI-based VP segmentation by designing an Adaptive Segmentation Network (ASN).

Approach: ASN uses adaptive convolution (AC) to dynamically adjust the kernel based on complementary context, facilitating the integration of contextual information. A spatial attention block selectively extracts relevant regions‘ features in each MRI sequence and fuses them. 

Results: ASN's effectiveness is validated by segmenting the VP in MR images from two MRI sequences. It surpasses state-of-the-art techniques in VP segmentation.

Impact: The introduction of ASN, a new multi-parametric MR images segmentation approach, demonstrates superior performance in visual pathway (VP) segmentation in MR images, surpassing existing state-of-the-art techniques. This novel method effectively incorporates context information, leading to improved segmentation performance.

Keywords: Data Acquisition, Software Tools, Protocol Compliance

Motivation: Large MRI datasets from multiple sites are not monitored for protocol compliance and dataset integrity. 

Goal(s): We previously demonstrated the pervasiveness of protocol non-compliance in MR datasets using our open source tool mrQA. We aim to produce deeper insights with vertical audit and analyze the common patterns of non-compliance.

Approach: We processed the large and open ABCD study verifying relationships between sequences in their protocol.

Results: We observed issues on non-compliance in coil, shim setting, and pixel spacing. We also observed significant disparities across vendors, scanners and sites. This underscores the necessity for  tools such as mrQA that can identify non-compliance across vendors/sites.

Impact: Non-compliance in acquisition parameters is a pervasive problem in MR datasets. It is impractical to “hope” for protocol compliance across sites, and scanners. Our tool, mrQA can enable researchers to continuously monitor and identify non-compliant scans in a practical manner. 

Keywords: Artifacts, Data Acquisition, Image Reconstruction

Motivation: Multi-shot overlapping-echo detachment imaging (msh-MOLED), a msh-EPI-based quantitative MR sequence, quantifies tissue T₂ rapidly without the need of separately acquiring images with different TEs, but its results could be contaminated by flow-induced inter-shot phase variations.

Goal(s): To implement an instantaneous referenceless flow-artifact correction for msh-MOLED.

Approach: Flow-related features were added to the training data, and the trained network fulfilled T₂ mapping free from flow artifacts without dear computational costs or additional reference data.

Results: After correction, the Pearson’s correlation coefficient/mean absolute error was changed from 0.6332/6.5328 (uncorrected) to 0.8808/2.7623 (corrected).

Impact: The proposed correction could be used to retain the mapping accuracy of msh-MOLED regardless of shot numbers, or to refine the reference data in high-spatial-resolution diffusion mapping potentially.

Keywords: Artifacts, Brain, Quality, Deep learning, Quality Assessment

Motivation: Deep learning (DL) outperforms conventional machine learning (ML) that relies on handcrafted feature-based in many vision tasks, but its superiority in assessing brain MRI image quality for new sites/scanners is unclear.

Goal(s): Compare DL and conventional ML  for quality assessment of brain MRI images from new sites/scanners.

Approach: One popular and widely accepted DL and one conventional ML method are  evaluated on a multi-site dataset using leave-one-site-out approach using a binary quality label (good/bad). 

Results: Averaged balanced accuracy (BA) for the DL and conventional ML approaches are comparably poor (0.60+-0.12 and 0.54+-0.12, respectively) and does not exceed 0.76, suggesting room for improvement. 

Impact: Widespread adoption of automated quality assessment of brain MRI images is limited by a lack of generalizability. By comparing popular DL and conventional ML approaches, we find comparable but limited generalizability. This underscores the need for  future algorithm development. 

Keywords: Quantitative Imaging, Quantitative Susceptibility mapping, Susceptibility Tensor Imaging

Motivation: To overcome the clinical limitations of susceptibility tensor imaging (STI) due to the requirement for multiple head orientations.

Goal(s): Develop a method to isolate χ₁₃ and χ₂₃ of the magnetic susceptibility tensor, from a single head orientation, enhancing the clinical viability of STI.

Approach: Employing a deep learning-based autoencoder, calibrated via STI and optimized for each dataset to separate the tensor components without the need for training or data rotation.

Results: The method successfully extracted χ₁₃ and χ₂₃ components comparable to the gold standard multi-orientation STI, showing potential for improved brain tissue characterization in conditions like multiple sclerosis.

Impact: We present a simplified STI approach, extracting critical tensor components from a single orientation scan. The new technique allows to assess structural tissue integrity, particularly in white matter. Requiring only a single orientation renders it clinically feasible.

Keywords: Motion Correction, Motion Correction, 3D radial, projection moment, center-of-mass

Motivation: Traditional projection moment analysis in 3D radial MRI failed to get specific rigid-body motion parameters with stationary multichannel coils.

Goal(s): Our goal was to develop a method to extract rigid-body motion parameters directly through projection moment analysis.

Approach: A PCA-based coil compression, together with projection information from different channels were used to estimate rigid-body motion parameters. A recursive least-squares model was used to recursively estimate motion parameters for every single spoke. Simulation and scanning of moving object were performed to demonstrate its capability in brain scan.

Results: The proposed method can correct motion in brain successfully and quickly.

Impact: The proposed method provides an easy, robust, and time-efficient tool for motion correction in brain MRI, which may benefit clinical diagnosis of uncooperative patients like children, in addition to many other applications including extremity MRI.

Keywords: Quantitative Imaging, Quantitative Susceptibility mapping, Susceptibility Inversion,Deep Learning

Motivation: Quantitative susceptibility mapping (QSM) estimates the spatial distribution of tissue susceptibility by solving a challenging ill-posed dipole inversion problem, which heavily affects the accuracy of tissue susceptibility quantification.

Goal(s): To generate high-quality QSM images.

Approach: In this study, we present a deep learning method for susceptibility inversion that utilizes a nonlinear susceptibility inversion model, NSIDL.Our approach integrated the Proximal Gradient Descent (PGD)^[1] algorithm and embedding the physical model in the network. 

Results: NSIDL was compared to traditional and deep learning methods, and it was found that NSIDL can effectively suppress streaking artifacts, mitigate noise amplification, and prevent excessive smoothing.

Impact: This study introduced the NSIDL deep learning method, which improved the accuracy of tissue magnetic sensitivity quantification. The improvement of QSM performance can help clinical doctors make more informed decisions based on reliable sensitivity measurements.