Number Of Diffusion Gradient Directions Research Articles

FlexDTI: flexible diffusion gradient encoding scheme-based highly efficient diffusion tensor imaging using deep learning

Objective. Most deep neural network-based diffusion tensor imaging methods require the diffusion gradients’ number and directions in the data to be reconstructed to match those in the training data. This work aims to develop and evaluate a novel dynamic-convolution-based method called FlexDTI for highly efficient diffusion tensor reconstruction with flexible diffusion encoding gradient scheme. Approach. FlexDTI was developed to achieve high-quality DTI parametric mapping with flexible number and directions of diffusion encoding gradients. The method used dynamic convolution kernels to embed diffusion gradient direction information into feature maps of the corresponding diffusion signal. Furthermore, it realized the generalization of a flexible number of diffusion gradient directions by setting the maximum number of input channels of the network. The network was trained and tested using datasets from the Human Connectome Project and local hospitals. Results from FlexDTI and other advanced tensor parameter estimation methods were compared. Main results. Compared to other methods, FlexDTI successfully achieves high-quality diffusion tensor-derived parameters even if the number and directions of diffusion encoding gradients change. It reduces normalized root mean squared error by about 50% on fractional anisotropy and 15% on mean diffusivity, compared with the state-of-the-art deep learning method with flexible diffusion encoding gradient scheme. Significance. FlexDTI can well learn diffusion gradient direction information to achieve generalized DTI reconstruction with flexible diffusion gradient scheme. Both flexibility and reconstruction quality can be taken into account in this network.

Node2Node: Self-Supervised Cardiac Diffusion Tensor Image Denoising Method

Although the existing cardiac diffusion tensor imaging (DTI) denoising methods have achieved promising results, most of them are dependent on the number of diffusion gradient directions, noise distributions, and noise levels. To address these issues, we propose a novel self-supervised cardiac DTI denoising network, Node2Node, which firstly expresses the diffusion-weighted (DW) image volumes along different directions as a graph, then the graph framelet transform (GFT) is implemented to map the DW signals into the GFT coefficients at different spectral bands, allowing us to accurately match the DW image pairs. After that, using the matched image pairs as input and target, a ResNet-like network is used to denoise in a self-supervised manner. In addition, a novel edge-aware loss based on pooling operation is proposed to retain the edge. Through comparison with several state-of-the-art methods on synthetic, ex vivo porcine, and in vivo human cardiac DTI datasets, we showed that the root mean square error (RMSE) of DW images and the average angular error (AAE) of fiber orientations obtained using Node2Node are the smallest, improved by 47.5% and 23.7%, respectively, on the synthetic dataset, demonstrating that Node2Node is not sensitive to the properties of the dataset.

Investigation of anisotropic fishing line-based phantom as tool in quality control of diffusion tensor imaging.

This work proposes a low-cost, fishing line-based phantom for quality control of diffusion tensor imaging (DTI). The device was applied to investigate the relationship between DTI indexes (DTIi) and imaging acquisition parameters. A Dyneema® fishing line phantom was built with fiber bundles of different thicknesses. DTI acquisitions were performed in a 3T magnetic resonance imaging scanner using an 8-channel and a 32-channel head coil. For each coil, the following acquisition parameters were changed, one at a time: diffusion sensitivity factor (b value), echo time, sensitivity encoding, voxel size, number of signal averages, and number of diffusion gradient directions (NDGD). DTIi including fractional anisotropy, relative anisotropy (RA), linear anisotropy (CL), and planar anisotropy (CP) were calculated for each image; the data were analyzed using the coefficient of variation (CV) and distributions of DTIi values. The 32-channel head coil presented higher CV values for the DTIi RA, CL, and CP when voxel size was changed. Using the phantom, dependences between diffusion-related parameters (b value and NDGD) and DTIi were also observed; the majority of these were for the smaller thickness fiber bundles. The device proved to be useful for the verification of the DTI performance over time.

Analytical performance bounds for multi-tensor diffusion-MRI

PurposeTo examine the effects of MR acquisition parameters on brain white matter fiber orientation estimation and parameter of clinical interest in crossing fiber areas based on the Multi-Tensor Model (MTM). Material and methodsWe compute the Cramér–Rao Bound (CRB) for the MTM and the parameter of clinical interest such as the Fractional Anisotropy (FA) and the dominant fiber orientations, assuming that the diffusion MRI data are recorded by a multi-coil, multi-shell acquisition system. Considering the sum-of-squares method for the reconstructed magnitude image, we introduce an approximate closed-form formula for Fisher Information Matrix that has the simplicity and easy interpretation advantages. In addition, we propose to generalize the FA and the mean diffusivity to the multi-tensor model. ResultsWe show the application of the CRB to reduce the scan time while preserving a good estimation precision. We provide results showing how the increase of the number of acquisition coils compensates the decrease of the number of diffusion gradient directions. We analyze the impact of the b-value and the Signal-to-Noise Ratio (SNR). The analysis shows that the estimation error variance decreases with a quadratic rate with the SNR, and that the optimum b-values are not unique but depend on the target parameter, the context, and eventually the target cost function. ConclusionIn this study we highlight the importance of choosing the appropriate acquisition parameters especially when dealing with crossing fiber areas. We also provide a methodology for the optimal tuning of these parameters using the CRB.

Distribution-aware estimation of the minimum achievable uncertainty in diffusion-tensor imaging (DTI).

Diffusion tensor imaging (DTI) provides exquisite sensitivity to structural and microstructural characteristics of brain tissue, and is routinely employed in advanced neuroimaging applications. DTI is commonly performed using intrinsically noisy echo-planar imaging techniques and poses high demands both on scanner performance and on in-scanner subject time, which in turn is directly related to the number of diffusion-weighting direction one requires. While DTI-derived indices such as fractional anisotropy (FA), diffusion tensor trace and anisotropy mode have proven extremely useful in characterizing disease-related aberrations, their estimation is commonly performed using fitting routines that do not properly take into account MRI noise distribution. In this paper, we present a distribution-aware maximum likelihood tensor estimation framework which also allows, for the first time, separate local noise estimation in both diffusion weighted and reference images. We validate our framework using multiple water phantom diffusion weighted acquisitions, and demonstrate its feasibility in human data. We then employ our framework within Monte Carlo simulations to show how the minimum achievable uncertainty attainable in DTI depends on signal-to-noise ratio (SNR) and number of diffusion gradient directions, demonstrating that these dependencies could be recast into simple power laws which may serve as guidelines for application-specific DTI protocol design.

Cervical Spinal Cord DTI Is Improved by Reduced FOV with Specific Balance between the Number of Diffusion Gradient Directions and Averages.

Reduced-FOV DTI is promising for exploring the cervical spinal cord, but the optimal set of parameters needs to be clarified. We hypothesized that the number of excitations should be favored over the number of diffusion gradient directions regarding the strong orientation of the cord in a single rostrocaudal axis. Fifteen healthy individuals underwent cervical spinal cord MR imaging at 3T, including an anatomic 3D-Multi-Echo Recombined Gradient Echo, high-resolution full-FOV DTI with a NEX of 3 and 20 diffusion gradient directions and 5 sets of reduced-FOV DTIs differently balanced in terms of NEX/number of diffusion gradient directions: (NEX/number of diffusion gradient directions = 3/20, 5/16, 7/12, 9/9, and 12/6). Each DTI sequence lasted 4 minutes 30 seconds, an acceptable duration, to cover C1-C4 in the axial plane. Fractional anisotropy maps and tractograms were reconstructed. Qualitatively, 2 radiologists rated the DTI sets blinded to the sequence. Quantitatively, we compared distortions, SNR, variance of fractional anisotropy values, and numbers of detected fibers. Qualitatively, reduced-FOV DTI sequences with a NEX of ≥5 were significantly better rated than the full-FOV DTI and the reduced-FOV DTI with low NEX (N = 3) and a high number of diffusion gradient directions (D = 20). Quantitatively, the best trade-off was reached by the reduced-FOV DTI with a NEX of 9 and 9 diffusion gradient directions, which provided significantly fewer artifacts, higher SNR on trace at b = 750 s/mm2 and an increased number of fibers tracked while maintaining similar fractional anisotropy values and dispersion. Optimized reduced-FOV DTI improves spinal cord imaging. The best compromise was obtained with a NEX of 9 and 9 diffusion gradient directions, which emphasizes the need for increasing the NEX at the expense of the number of diffusion gradient directions for spinal cord DTI contrary to brain DTI.

A Bayesian approach to distinguishing interdigitated tongue muscles from limited diffusion magnetic resonance imaging

The tongue is a critical organ for a variety of functions, including swallowing, respiration, and speech. It contains intrinsic and extrinsic muscles that play an important role in changing its shape and position. Diffusion tensor imaging (DTI) has been used to reconstruct tongue muscle fiber tracts. However, previous studies have been unable to reconstruct the crossing fibers that occur where the tongue muscles interdigitate, which is a large percentage of the tongue volume. To resolve crossing fibers, multi-tensor models on DTI and more advanced imaging modalities, such as high angular resolution diffusion imaging (HARDI) and diffusion spectrum imaging (DSI), have been proposed. However, because of the involuntary nature of swallowing, there is insufficient time to acquire a sufficient number of diffusion gradient directions to resolve crossing fibers while the in vivo tongue is in a fixed position. In this work, we address the challenge of distinguishing interdigitated tongue muscles from limited diffusion magnetic resonance imaging by using a multi-tensor model with a fixed tensor basis and incorporating prior directional knowledge. The prior directional knowledge provides information on likely fiber directions at each voxel, and is computed with anatomical knowledge of tongue muscles. The fiber directions are estimated within a maximum a posteriori (MAP) framework, and the resulting objective function is solved using a noise-aware weighted ℓ1-norm minimization algorithm. Experiments were performed on a digital crossing phantom and in vivo tongue diffusion data including three control subjects and four patients with glossectomies. On the digital phantom, effects of parameters, noise, and prior direction accuracy were studied, and parameter settings for real data were determined. The results on the in vivo data demonstrate that the proposed method is able to resolve interdigitated tongue muscles with limited gradient directions. The distributions of the computed fiber directions in both the controls and the patients were also compared, suggesting a potential clinical use for this imaging and image analysis methodology.

Effect of increasing diffusion gradient direction number on diffusion tensor imaging fiber tracking in the human brain.

ObjectiveTo assess the effects of varying the number of diffusion gradient directions (NDGDs) on diffusion tensor fiber tracking (FT) in human brain white matter using tract characteristics.Materials and MethodsTwelve normal volunteers underwent diffusion tensor imaging (DTI) scanning with NDGDs of 6, 11, 15, 21, and 31 orientations. Three fiber tract groups, including the splenium of the corpus callosum (CC), the entire CC, and the full brain tract, were reconstructed by deterministic DTI-FT. Tract architecture was first qualitatively evaluated by visual observation. Six quantitative tract characteristics, including the number of fibers (NF), average length (AL), fractional anisotropy (FA), relative anisotropy (RA), mean diffusivity (MD), and volume ratio (VR) were measured for the splenium of the CC at the tract branch level, for the entire CC at tract level, and for the full brain tract at the whole brain level. Visual results and those of NF, AL, FA, RA, MD, and VR were compared among the five different NDGDs.ResultsThe DTI-FT with NDGD of 11, 15, 21, and 31 orientations gave better tracking results compared with NDGD of 6 after the visual evaluation. NF, FA, RA, MD, and VR values with NDGD of six were significantly greater (smallest p = 0.001 to largest p = 0.042) than those with four other NDGDs (11, 15, 21, or 31 orientations), whereas AL measured with NDGD of six was significantly smaller (smallest p = 0.001 to largest p = 0.041) than with four other NDGDs (11, 15, 21, or 31 orientations). No significant differences were observed in the results among the four NDGD groups of 11, 15, 21, and 31 directions (smallest p = 0.059 to largest p = 1.000).ConclusionThe main fiber tracts were detected with NDGD of six orientations; however, the use of larger NDGD (≥ 11 orientations) could provide improved tract characteristics at the expense of longer scanning time.

Effects of MR parameter changes on the quantification of diffusion anisotropy and apparent diffusion coefficient in diffusion tensor imaging: evaluation using a diffusional anisotropic phantom.

ObjectiveTo validate the usefulness of a diffusional anisotropic capillary array phantom and to investigate the effects of diffusion tensor imaging (DTI) parameter changes on diffusion fractional anisotropy (FA) and apparent diffusion coefficient (ADC) using the phantom.Materials and MethodsDiffusion tensor imaging of a capillary array phantom was performed with imaging parameter changes, including voxel size, number of sensitivity encoding (SENSE) factor, echo time (TE), number of signal acquisitions, b-value, and number of diffusion gradient directions (NDGD), one-at-a-time in a stepwise-incremental fashion. We repeated the entire series of DTI scans thrice. The coefficients of variation (CoV) were evaluated for FA and ADC, and the correlation between each MR imaging parameter and the corresponding FA and ADC was evaluated using Spearman's correlation analysis.ResultsThe capillary array phantom CoVs of FA and ADC were 7.1% and 2.4%, respectively. There were significant correlations between FA and SENSE factor, TE, b-value, and NDGD, as well as significant correlations between ADC and SENSE factor, TE, and b-value.ConclusionA capillary array phantom enables repeated measurements of FA and ADC. Both FA and ADC can vary when certain parameters are changed during diffusion experiments. We suggest that the capillary array phantom can be used for quality control in longitudinal or multicenter clinical studies.

Optimization of the Parameters for Diffusion Tensor Magnetic Resonance Imaging Data Acquisition for Breast Fiber Tractography at 1.5 T

IntroductionDiffusion tensor MRI has emerged as a promising tool for the analysis of the microscopic properties of tissues. Optimizing image acquisition parameters is essential for producing high-quality DTI. This study aimed to optimize the parameters for DTI data acquisition for breast fiber tractography at 1.5 T. Patients and MethodsA total of 21 healthy volunteers received breast DTI scanning using an ASSET-based EPI technique operated under different parameters including b value, the number of diffusion gradient directions, and spatial resolution. The images were analyzed for signal-to-noise, signal intensity ratio, mean number and length of reconstructive fiber tracts, and fractional anisotropy value. ResultsThe optimal acquisition parameters at 1.5 T for breast DT-MRI fiber tractography were determined as follows: axial 31 direction, b = 600 seconds per mm2, matrix 128 × 128 with slice thickness of 3 mm. ConclusionThe optimization of data acquisition parameters could improve the quality of breast DT-MRI images and assist fiber tractography at 1.5 T.

The effect of gradient sampling schemes on diffusion metrics derived from probabilistic analysis and tract-based spatial statistics

PurposeThe purpose was to systematically evaluate the effect of diffusion gradient encoding scheme on estimated fractional anisotropy (FA), mean diffusivity (MD) and the voxel-wise probability of identifying crossing fibers in the brain. Materials and MethodsEight healthy volunteers (mean age 26.5±1.3 years, 5 males, 3 females) were imaged using a Spin-Echo Echo-Planar-Imaging sequence acquired with two signal averages [number of signals averaged (NSA)], 127 diffusion directions, and b-values of 750 s/mm2 and 1500 s/mm2. The number of diffusion gradient directions (Nd) was reduced from the original value whilst maintaining a homogeneous gradient distribution enabling direct comparison of subsampled data sets with Nd=15, 28, 43, 84, 112 and 127. FA and MD maps were generated and analyzed using tract-based spatial statistics. Effect of Nd on estimated FA and MD was tested with voxel-wise statistics in 13 regions of interest. The number of voxels supporting two fiber populations (NV2) at different Nd values was estimated using Bayesian estimation of diffusion parameters. ResultsLow FA values decreased significantly with increasing Nd and with increasing NSA. MD was only marginally sensitive to Nd and NSA. NV2 increased significantly with Nd but not with NSA. Thus, we conclude that accurate estimation of standard diffusion metrics FA and MD is mainly dependent on the signal-to-noise ratio (SNR), whereas the ability to differentiate multiple fiber populations requires a high diffusion sampling density.

Dependence of the fractional anisotropy in cervical spine from the number of diffusion gradients, repeated acquisition and voxel size

The aim of this study is to investigate the consequences of using different gradient schemes, number of repeated measurements and voxel size on the fractional anisotropy (FA) value in a diffusion tensor imaging (DTI) sequence on the cervical tract of the spinal cord. Twenty healthy volunteers underwent a total of 86 DTI axial acquisitions performed by using different voxel size and number of diffusion gradient directions (NDGDs). Three different diffusion gradient schemes were applied, named 6, 15 and 32 according to the NDGD. Furthermore, some acquisitions were repeated to investigate the effects of image averaging on FA value. Our results indicate that the FA value in the cervical spinal cord decreases when increasing the NDGD for a fixed spatial resolution, or when identical acquisitions are repeated, thus, increasing the acquisition time. This effect is observed in all subjects without exceptions, and the differences result statistically significant: the average FA obtained from 6, 15 and 32 NDGD is 0.84 (range, 0.82-0.87), 0.75 (range, 0.68-0.80) and 0.70 (range, 0.65-0.77), respectively, for isotropic 8 mm(3) voxel size. When varying the spatial resolution in a volume range of 2 to 8 mm(3) for a fixed NDGD (6 or 15), the differences in FA values are smaller albeit still statistically significant: the smaller the voxel, the larger the FA. No significant dependence of the FA value from the spatial resolution is observed in the 32 NDGD acquisitions in the studied volume range. In conclusion, our results indicate that the value of the FA in the cervical tract of the spinal cord vary with regularity in intrasubject acquisitions when modifying the NDGD and when repeated acquisitions are used; these observations confirm that the signal-to-noise ratio introduces a systematic error in FA measurements that does not allow simple comparison of quantitative results obtained from separated studies.

Evaluation of measurement uncertainties in human diffusion tensor imaging (DTI)‐derived parameters and optimization of clinical DTI protocols with a wild bootstrap analysis

To quantify measurement uncertainties of fractional anisotropy, mean diffusivity, and principal eigenvector orientations in human diffusion tensor imaging (DTI) data acquired with common clinical protocols using a wild bootstrap analysis, and to establish optimal scan protocols for clinical DTI acquisitions. A group of 13 healthy volunteers were scanned using three commonly used DTI protocols with similar total scan times. Two important parameters-the number of unique diffusion gradient directions (NUDG) and the ratio of the total number of diffusion-weighted (DW) images to the total number of non-DW images (DTIR)-were analyzed in order to investigate their combined effects on uncertainties of DTI-derived parameters, using results from both the Monte Carlo simulation and the wild bootstrap analysis of uncertainties in human DTI data. The wild bootstrap analysis showed that uncertainties in human DTI data are significantly affected by both NUDG and DTIR in many brain regions. These results agree with previous predictions based on error-propagations as well as results from simulations. Our results demonstrate that within a clinically feasible DTI scan time of about 10 minutes, a protocol with number of diffusion gradient directions close to 30 provides nearly optimal measurement results when combined with a ratio of the total number of DW images over non-DW images equal to six. Wild bootstrap can serve as a useful tool to quantify the measurement uncertainty from human DTI data.

Histogram Analysis of Noise Performance on Fractional Anisotropy Brain MR Image with Different Diffusion Gradient Numbers

Purpose: We wished to analyze, qualitatively and quantitatively, the noise performance of fractional anisotropy brain images along with the different diffusion gradient numbers by using the histogram method. Materials and Methods: Diffusion tensor images were acquired using a 3.0 T MR scanner from ten normal volunteers who had no neurological symptoms. The single-shot spin-echo EPI with a Stejskal-Tanner type diffusion gradient scheme was employed for the diffusion tensor measurement. With a b-valuee of 1000 s/, the diffusion tensor images were obtained for 6, 11, 23, 35 and 47 diffusion gradient directions. FA images were generated for each DTI scheme. The histograms were then obtained at selected ROIs for the anatomical structures on the FA image. At the same ROI location, the mean FA value and the standard deviation of the mean FA value were calculated. Results: The quality of the FA image was improved as the number of diffusion gradient directions increased by showing better contrast between the WM and GM. The histogram showed that the variance of FA values was reduced as the number of diffusion gradient directions increased. This histogram analysis was in good agreement with the result obtained using quantitative analysis. Conclusion: The image quality of the FA map was significantly improved as the number of diffusion gradient directions increased. The histogram analysis well demonstrated that the improvement in the FA images resulted from the reduction in the variance of the FA values included in the ROI.