Magnetic Resonance Materials in Physics, Biology and Medicine最新文献

Purpose: To enable accelerated Bloch simulations by enhancing the open-source multi-purpose MRI simulation tool JEMRIS with graphic processing units (GPU) parallelization.

Methods: A GPU-compatible version of JEMRIS was built by shifting the computationally expensive parallelizable processes to the GPU to benefit from heterogeneous computing and by adding asynchronous communication and mixed precision support. With key classes reimplemented in CUDA C++, the developed GPU-JEMRIS framework was tested on simulations of common MRI artifacts in numerical phantoms. The accuracy and performance of the GPU-parallelized JEMRIS simulator were benchmarked against the CPU-parallelized JEMRIS and GPU-enabled KomaMRI.jl simulators. Additionally, an example of liver fat quantification errors due to respiratory motion artifacts was simulated in a multi-echo gradient echo (MEGRE) acquisition.

Results: The GPU-accelerated JEMRIS achieved speed-up factors 3-12 and 7-65 using double and single precision numerical integrators, respectively, when compared to the parallelized CPU implementation in the investigated numerical phantom scenarios. While double precision GPU simulations negligibly differ (<0.1% NRMSE) from double precision CPU simulations, the single precision simulations still present small errors of up to 1% k-space signal NRMSE. The developed a GPU extension enabled computationally demanding motion simulations with a multi-species abdominal phantom and a MEGRE sequence, showing significant and spatially varying fat fraction bias in the presence of motion.

Conclusion: By solving the Bloch equations in parallel on device, accelerated Bloch simulations can be performed on any GPU-equipped device with CUDA support using the developed GPU-JEMRIS. This would enable further insights into more realistic large spin pool MR simulations such as experiments with large multi-dimensional phantoms, multiple chemical species and dynamic effects.

Purpose: Cardiac magnetic resonance fingerprinting (cMRF) is a powerful quantitative imaging technique that provides multi-parametric diagnostic information. Here, we introduce an open-source framework for cardiac MRF including open-source pulse sequences, image reconstruction, and parameter estimation tools that are needed for the processing of the data.

Methods: A 2D cMRF sequence with a variable-density spiral readout is implemented using the open-source and vendor-agnostic sequence format Pulseq. Cardiac triggering is used to synchronize acquisition with the rest period of the heart. $T_1$ inversion and $T_2$ preparation pulses are added to ensure accurate parameter estimation. Data acquisition is carried out over 15 heartbeats. The images showing the signal changes over time are reconstructed and matched to a pre-calculated signal dictionary. In addition to the cMRF sequence, spin-echo reference sequences for quality control in phantoms are provided. The method is evaluated in phantom experiments using a T1MES phantom on four different scanners. In vivo experiments were performed to compare the open-source cMRF sequence with a vendor-specific cMRF sequence and clinical sequences used for $T_1$ and $T_2$ mapping of the heart. Three volunteers were imaged on two different scanners.

Results: The error of $T_1$ and $T_2$ over all tissue types present in the T1MES phantom was comparable between all four scanners and on average 4.50 ± 2.48%. $T_1$ and $T_2$ maps obtained in vivo were comparable between the open-source and vendor-specific implementation of cMRF.

Conclusion: The proposed open-source cMRF implementation enables accurate parameter estimation across multiple different scanners. Sequence files, image reconstruction, and parameter estimation scripts are available for reproducible quantitative MRI.

Objective: To validate the automated analysis of magnetic resonance imaging (MRI) diffusion phantoms with an updated version of the magnetic resonance biomarker assessment software (MR-BIAS), an open-source tool initially developed for the analysis of MRI relaxometry phantoms.

Materials and methods: The updated MR-BIAS was validated against two published diffusion weighted MRI datasets: (i) a single-site study (n = 48) was used for validation of apparent diffusion coefficients (ADC) and to identify optimal region of interest (ROI) selection, and (ii) a multi-centre multi-vendor study including diffusion imaging from a shared benchmark protocol (n = 49) and site-specific protocols (n = 43). ADC analysis compared both datasets with ROIs manually matched to the original studies, and with automatically detected optimal ROIs.

Results: MR-BIAS ADC values were statistically equivalent (p < 0.05) to original studies within tolerances (manual ROI, automatic ROI) for the single-site study (± 0.01, ± 6 μm²/s) and for the multi-vendor study for benchmark (± 4, ± 7 μm²/s) and site-specific (± 3, ± 6 μm²/s) protocols. The optimal ROI was a central cylinder (height = 10mm, diameter = 10mm). MR-BIAS ADC summary metrics were comparable to those of the original studies.

Discussion: MR-BIAS can automatically and accurately perform ADC analysis of diffusion phantoms, making the software suitable for the quality assurance of multi-centre studies of multi-parametric MRI.

Objective: To assess multi-site and multi-vendor accuracy, and intra- and inter-scanner variability of T1 and T2 measurements using the ISMRM/NIST System MRI phantom at room temperature.

Materials and methods: T1 and T2 measurements were acquired using standardized NIST protocols on 13 scanners (1.5 T and 3 T) from 3 vendors at 7 sites and compared with reference values at room temperature. Pearson's correlation (r) and accuracy error were used for comparison with reference values, while inter-scanner agreement was assessed using the coefficient of variation (CV%). Short-term reproducibility was evaluated using Bland-Altman plots and precision error. Generalized linear mixed models and post hoc tests (α = 0.05) were adopted to compare accuracy and precision across field strengths, vendors, and scanners. T2 measurements were corrected with StimFit toolbox for stimulated echo compensation.

Results: T1 and T2 measurements had excellent correlation with reference values at both field strengths. Stimfit significantly improved T2 accuracy in the renal range for 9 of 13 scanners. Short-term reproducibility (limits of agreement < 10%) and inter-scanner agreement were good (median CV < 7%) for both T1 and T2 values. Inter-scanner CV was < 5% in the renal range for both parameters.

Discussion: These findings support the need of scanner evaluation processes to ensure reliable T1-T2 measurements in multi-center MRI studies.

Objectives: By rapidly changing magnetic field strength between 0.2 and 200 mT during the pulse sequence Field-Cycling Imaging (FCI) makes it possible to identify and evaluate new quantitative markers of pathology derived from dispersion of spin-lattice relaxation rate (R₁) in vivo. The aim of this work was to determine the most effective approach to reliably estimate multi-field R₁ dispersion measurements in brain tissue using FCI.

Materials and methods: This repeatability study consisted of twenty participants with moderate or severe small vessel disease. Each participant underwent 3 T MRI and FCI scans, repeated 30 days apart. After R₁ maps were generated at 0.2, 2, 20, and 200 mT, co-registered tissue labels generated from 3 T MRI were used to extract tissue averaged values of R₁ dispersion from regions of white matter (WM) and WM hyperintensities (WMHs).

Results: The fitted model which yielded best overall image contrast between WM and WMH regions and R₁ dispersion model adherence was determined. Tissue averaged values of R₁ (0.2 mT) and R₁ dispersion slope exhibited Cohen's d effect sizes of 3.07 and 1.48, respectively, between regions of WM and WMH. The cohort study results were repeatable between study visits.

Discussion: Differences in R₁ measurements could repeatably be discerned between normal and abnormal appearing brain tissues.

Objective: This work presents an automated quality control (QC) system within quantitative MRI (qMRI) workflows. By leveraging the ISMRM/NIST quantitative MRI system phantom, we establish an open-source pipeline for rapid, repeatable, and accurate validation and stability tracking of sequence quantification performance across diverse clinical settings.

Materials and methods: A microservice-based QC system for automated vial segmentation from quantitative maps was developed and tested across various MRF acquisition and protocol designs, with reports generated and returned to the scanner in real time.

Results: The system demonstrated consistent and repeatable value segmentation and reporting, successfully extracted all 252 T1 and T2 vial samples tested. Values extracted from the same sequence were found to be repeatable with 0.09% ± 1.23% and - 0.26% ± 2.68% intersession error, respectively.

Discussion: By providing real-time quantification performance assessment, this easily deployable automated QC approach streamlines sequence validation and long-term performance monitoring, vital for the broader acceptance of qMRI as a standard component of clinical protocols.

Objectives: Consistent image quality and signal stability are crucial for neuroimaging, particularly fMRI studies that rely on detecting small BOLD signal changes. Regular MR system performance monitoring is essential, especially for longitudinal and multi-site studies. This work aims to establish a robust quality assurance (QA) protocol to enhance data comparability across days, scanner versions, vendors, and sites.

Materials and methods: We implemented an open-source, vendor-independent QA protocol using Pulseq for standardized data acquisition and ISMRMRD/Gadgetron for harmonized image reconstruction, accompanied by an automated post-processing pipeline to evaluate structural and temporal image quality. The protocol was thoroughly tested on three Siemens 3T scanners with different software versions at one site, and one GE 3T scanner at another site. The test was repeated on an fBIRN phantom for at least 4 days.

Results: The vendor-independent protocol produced image quality comparable to a closely matched vendor-based protocol. It showed similar day-to-day repeatability to the vendor-based protocol across the Siemens scanners and high inter-day repeatability on the GE scanner.

Conclusion: We successfully developed and implemented an open-source, vendor-independent QA protocol, accompanied by an automated post-processing pipeline. The results demonstrate the feasibility and repeatability of the protocol across different days, system versions, vendors, and sites.

Objective: AI-based MRI reconstruction techniques improve efficiency by reducing acquisition times whilst maintaining or improving image quality. Recent recommendations from professional bodies suggest centres should perform quality assessments on AI tools. However, monitoring long-term performance presents challenges, due to model drift or system updates. Radiologist-based assessments are resource-intensive and may be subjective, highlighting the need for efficient quality control (QC) measures. This study explores using image quality metrics (IQMs) to assess AI-based reconstructions.

Materials and methods: 58 patients undergoing standard-of-care rectal MRI were imaged using AI-based and conventional T2-weighted sequences. Paired and unpaired IQMs were calculated. Sensitivity of IQMs to detect retrospective perturbations in AI-based reconstructions was assessed using control charts, and statistical comparisons between the four MR systems in the evaluation were performed. Two radiologists evaluated the image quality of the perturbed images, giving an indication of their clinical relevance.

Results: Paired IQMs demonstrated sensitivity to changes in AI-reconstruction settings, identifying deviations outside ± 2 standard deviations of the reference dataset. Unpaired metrics showed less sensitivity. Paired IQMs showed no difference in performance between 1.5 T and 3 T systems (p > 0.99), whilst minor but significant (p < 0.0379) differences were noted for unpaired IQMs.

Discussion: IQMs are effective for QC of AI-based MR reconstructions, offering resource-efficient alternatives to repeated radiologist evaluations. Future work should expand this to other imaging applications and assess additional measures.

Quantitative MRI has been an active area of research for decades and has produced a huge range of approaches with enormous potential for patient benefit. In many cases, however, there are challenges with reproducibility which have hampered clinical translation. Quantitative MRI is a form of measurement and like any other form of measurement it requires a supporting metrological framework to be fully consistent and compatible with the international system of units. This means not just expressing results in terms of seconds, meters, etc., but demonstrating consistency to their internationally recognized definitions. Such a framework for MRI is not yet complete, but a considerable amount of work has been done internationally towards building one. This article describes the current state of the art for MRI metrology, including a detailed description of metrological principles and how they are relevant to fully quantitative MRI. It also undertakes a gap analysis of where we are versus where we need to be to support reproducibility in MRI. It focusses particularly on the role and activities of national measurement institutes across the globe, illustrating the genuinely international and collaborative nature of the field.

In quantitative susceptibility mapping (QSM), it is impossible to define an absolute reference for the reconstructed susceptibility values. Therefore, it has been suggested to use a relative reference, such as the mean susceptibility within an anatomical ROI. We investigated the theoretical basis of referencing, and what impact it may have on statistical ROI comparisons, particularly for clinical applications. We analysed a clinical epilepsy study and in-silico QSM reconstruction challenge data with various reference regions. The results are analysed as in a clinical study and resulting statistical variations are investigated from a theoretical point of view. We found that referencing has an impact on the significance of clinical findings. These effects may arise from a change in the precision of test statistics due to referencing. We also show potential biasing of results from referencing. Our findings suggest there may not be one "optimal" reference region, and care should always be taken with reference region selection depending on the specific pathology or cohort under investigation. Not explicitly referencing is less likely to lead to false positives than cherry picking a reference region to maximize statistically significant results. We encourage results to be published with their reference to facilitate future comparisons of datasets from different sources.