Voxel Dimensions Research Articles

Comparison of human and automatic segmentations of kidneys from CT images

A controlled observer study was conducted to compare a method for automatic image segmentation with conventional user-guided segmentation of right and left kidneys from planning computerized tomographic (CT) images. Deformable shape models called m-reps were used to automatically segment right and left kidneys from 12 target CT images, and the results were compared with careful manual segmentations performed by two human experts. M-rep models were trained based on manual segmentations from a collection of images that did not include the targets. Segmentation using m-reps began with interactive initialization to position the kidney model over the target kidney in the image data. Fully automatic segmentation proceeded through two stages at successively smaller spatial scales. At the first stage, a global similarity transformation of the kidney model was computed to position the model closer to the target kidney. The similarity transformation was followed by large-scale deformations based on principal geodesic analysis (PGA). During the second stage, the medial atoms comprising the m-rep model were deformed one by one. This procedure was iterated until no changes were observed. The transformations and deformations at both stages were driven by optimizing an objective function with two terms. One term penalized the currently deformed m-rep by an amount proportional to its deviation from the mean m-rep derived from PGA of the training segmentations. The second term computed a model-to-image match term based on the goodness of match of the trained intensity template for the currently deformed m-rep with the corresponding intensity data in the target image. Human and m-rep segmentations were compared using quantitative metrics provided in a toolset called Valmet. Metrics reported in this article include (1) percent volume overlap; (2) mean surface distance between two segmentations; and (3) maximum surface separation (Hausdorff distance). Averaged over all kidneys the mean surface separation was 0.12 cm, the mean Hausdorff distance was 0.99 cm, and the mean volume overlap for human segmentations was 88.8%. Between human and m-rep segmentations the mean surface separation was 0.18-0.19 cm, the mean Hausdorff distance was 1.14-1.25 cm, and the mean volume overlap was 82-83%. Overall in this study, the best m-rep kidney segmentations were at least as good as careful manual slice-by-slice segmentations performed by two experienced humans, and the worst performance was no worse than typical segmentations from our clinical setting. The mean surface separations for human-m-rep segmentations were slightly larger than for human-human segmentations but still in the subvoxel range, and volume overlap and maximum surface separation were slightly better for human-human comparisons. These results were expected because of experimental factors that favored comparison of the human-human segmentations. In particular, m-rep agreement with humans appears to have been limited largely by fundamental differences between manual slice-by-slice and true three-dimensional segmentation, imaging artifacts, image voxel dimensions, and the use of an m-rep model that produced a smooth surface across the renal pelvis.

Intrathoracic tumour motion estimation from CT imaging using the 3D optical flow method**Presented at The IASTED Second International Conference on Biomedical Engineering (BioMED 2004), Innsbruck, Austria, 16–18 February 2004.

The purpose of this work was to develop and validate an automated method for intrathoracic tumour motion estimation from breath-hold computed tomography (BH CT) imaging using the three-dimensional optical flow method (3D OFM). A modified 3D OFM algorithm provided 3D displacement vectors for each voxel which were used to map tumour voxels on expiration BH CT onto inspiration BH CT images. A thoracic phantom and simulated expiration/inspiration BH CT pairs were used for validation. The 3D OFM was applied to the measured inspiration and expiration BH CT images from one lung cancer and one oesophageal cancer patient. The resulting displacements were plotted in histogram format and analysed to provide insight regarding the tumour motion. The phantom tumour displacement was measured as 1.20 and 2.40 cm with full-width at tenth maximum (FWTM) for the distribution of displacement estimates of 0.008 and 0.006 cm, respectively. The maximum error of any single voxel's motion estimate was 1.1 mm along the z-dimension or approximately one-third of the z-dimension voxel size. The simulated BH CT pairs revealed an rms error of less than 0.25 mm. The displacement of the oesophageal tumours was nonuniform and up to 1.4 cm, this was a new finding. A lung tumour maximum displacement of 2.4 cm was found in the case evaluated. In conclusion, 3D OFM provided an accurate estimation of intrathoracic tumour motion, with estimated errors less than the voxel dimension in a simulated motion phantom study. Surprisingly, oesophageal tumour motion was large and nonuniform, with greatest motion occurring at the gastro-oesophageal junction.

Signal-to-noise ratio behavior of steady-state free precession.

Steady-state free precession (SSFP) is a rapid gradient-echo imaging technique that has recently gained popularity and is used in a variety of applications, including cardiac and real-time imaging, because of its high signal and favorable contrast between blood and myocardium. The purpose of this work was to examine the signal-to-noise ratio (SNR) behavior of images acquired with SSFP, and the dependence of SNR on imaging parameters such as TR, bandwidth, and image resolution, and the use of multi-echo sequences. In this work it is shown that the SNR of SSFP sequences is dependent only on pulse sequence efficiency, voxel dimensions, and relaxation parameters (T1 and T2). Notably, SNR is insensitive to bandwidth unless increases in bandwidth significantly decrease efficiency. Finally, we examined the relationship between pulse sequence performance (TR and efficiency) and gradient performance (maximum gradient strength and slew rate) for several imaging scenarios, including multi-echo sequences, to determine the optimum matching of maximum gradient strength and slew rate for gradient hardware designs. For standard modern gradient hardware (40 mT/m and 150 mT/m/ms), we found that the maximum gradient strength is more than adequate for the imaging resolution that is commonly encountered with rapid scouting (3 mm x 4 mm x 10 mm voxel). It is well matched for typical CINE and real-time cardiac imaging applications (1.5 mm x 2 mm x 6 mm voxel), and is inadequate for optimal matching with slew rate for high-resolution applications such as musculoskeletal imaging (0.5 x 0.8 x 3 mm voxel). For the lower-resolution methods, efficiency could be improved with higher slew rates; this provokes interest in designing methods for limiting dB/dt peripherally while achieving high switching rates in the imaging field of view. The use of multi-echo SSFP acquisitions leads to substantial improvements in sequence performance (i.e., increased efficiency and shorter TR).

A novel phantom and method for comprehensive 3-dimensional measurement and correction of geometric distortion in magnetic resonance imaging.

A phantom that can be used for mapping geometric distortion in magnetic resonance imaging (MRI) is described. This phantom provides an array of densely distributed control points in three-dimensional (3D) space. These points form the basis of a comprehensive measurement method to correct for geometric distortion in MR images arising principally from gradient field non-linearity and magnet field inhomogeneity. The phantom was designed based on the concept that a point in space can be defined using three orthogonal planes. This novel design approach allows for as many control points as desired. Employing this novel design, a highly accurate method has been developed that enables the positions of the control points to be measured to sub-voxel accuracy. The phantom described in this paper was constructed to fit into a body coil of a MRI scanner, (external dimensions of the phantom were: 310 mm x 310 mm x 310 mm), and it contained 10,830 control points. With this phantom, the mean errors in the measured coordinates of the control points were on the order of 0.1 mm or less, which were less than one tenth of the voxel's dimensions of the phantom image. The calculated three-dimensional distortion map, i.e., the differences between the image positions and true positions of the control points, can then be used to compensate for geometric distortion for a full image restoration. It is anticipated that this novel method will have an impact on the applicability of MRI in both clinical and research settings, especially in areas where geometric accuracy is highly required, such as in MR neuro-imaging.

Functional coupling of simultaneous electrical and metabolic activity in the human brain.

The relationships between brain electrical and metabolic activity are being uncovered currently in animal models using invasive methods; however, in the human brain this relationship remains not well understood. In particular, the relationship between noninvasive measurements of electrical activity and metabolism remains largely undefined. To understand better these relations, cerebral activity was measured simultaneously with electroencephalography (EEG) and positron emission tomography using [(18)f]-fluoro-2-deoxy-D-glucose (PET-FDG) in 12 normal human subjects during rest. Intracerebral distributions of current density were estimated, yielding tomographic maps for seven standard EEG frequency bands. The PET and EEG data were registered to the same space and voxel dimensions, and correlational maps were created on a voxel-by-voxel basis across all subjects. For each band, significant positive and negative correlations were found that are generally consistent with extant understanding of EEG band power function. With increasing EEG frequency, there was an increase in the number of positively correlated voxels, whereas the lower alpha band (8.5-10.0 Hz) was associated with the highest number of negative correlations. This work presents a method for comparing EEG signals with other more traditionally tomographic functional imaging data on a 3-D basis. This method will be useful in the future when it is applied to functional imaging methods with faster time resolution, such as short half-life PET blood flow tracers and functional magnetic resonance imaging.

In vivo quantitative analysis of the effect of hydration (immersion and Vaseline treatment) in skin layers using high‐resolution MRI and magnetisation transfer contrast*

Many claims are made as to the efficacy of topical preparations in moisturising the skin, yet most of these claims cannot be substantiated by scientific study for the skin layers beneath the stratum corneum, and yield no information on the remainder of the epidermis and dermis. This argues for an in vivo quantitative method for measuring the effect of water loading extended to various layers of the skin. Detailed high-resolution in vivo MRI studies of hydration and dehydration of finger pad skin layers were conducted on one normal subject using two moisturisation methods (topical white soft paraffin (Vaseline) and water immersion). The dehydration study was carried out immediately following removal from prolonged skin moisturisation. Inter-individual variability for skin hydration (group study) was studied in seven healthy volunteers at 0 and 7 h hydration with Vaseline. Location dependence in skin hydration was investigated on the same subject by looking into the hydration of forearm and finger pad skin. System stability and measurement reproducibility was verified through a detailed phantom study. Images of normal and hydrated human skin were obtained in vivo at voxel dimensions of 50 micromx150 micromx1000 microm. The effect of hydration and dehydration as a function of exposure to moisturiser (i.e. water and Vaseline) on the image signal intensity, observed T1, and interaction of free and bound water in specific tissues were identified and correlated with existing physiological knowledge. Swelling of stratum corneum due to hydration was expressed as an in vivo model of tissue hydration. Results of the dehydration study showed that the changes due to the previous hydration of the skin are reversible for all skin layers. For both moisturisation methods (i.e. Vaseline and skin bathing), the effects of hydration and dehydration on the skin were similar. The trends of the MRI parameters for finger pad and arm skin were similar. The group study showed low inter-subject variability of hydration on stratum corneum and epidermis.

Monte Carlo simulation for dose distribution calculations in a CT-based phantom at the Portuguese gamma irradiation facility

In preview works the Portuguese Gamma Irradiation Facility, UTR, has been simulated using the MCNP code and the product to be irradiated has been drawn using the boolean operators with the MCNP surfaces. However, sometimes the product to be irradiated could have an irregular shape. The paper describes an alternative way for drawing the corresponding volume based on CT image data in a format of a 3D matrix of voxels. This data are read by a specific code called SCMS which transforms it into a MCNP input file. The dimensions of each MCNP voxel depend on the number of elements in the CT-based matrix. Additionally, the new approach allows one to know dose distributions anywhere without extra definitions of surfaces or volumes. Experimental dose measurements were carried out using Amber Perspex dosimeters. This work presents the results of MCNP simulations using both modeling modes – the standard mode and the voxel mode.

The application accuracy of the Pathfinder neurosurgical robot

We have developed a new technique to measure the overall application accuracy of image-guided stereotactic systems. It has been used to measure the accuracy of the Pathfinder neurosurgical robot, which uses a robot-mounted camera to triangulate fiducial skin markers, virtually eliminating line-of-sight problems experienced by other optical localisation methods. A phantom consisting of plastic spheres on stalks was used. The targets are the centre of the spheres. Points on a sphere's surface were automatically selected on CT images, from which the centre of the sphere was calculated in image space to greater precision than the CT voxel dimensions. A depth gauge was attached to Pathfinder in place of a biopsy needle and driven to the target as defined on CT. Pathfinder then moved to several points over the sphere, moving perpendicular to the trajectory in a grid pattern. By considering the depth gauge reading, 3D points on the sphere's surface were measured in robot space and the centre of the sphere was found. The difference between the actual and planned targets yields the overall accuracy. Preliminary experiments show that the application accuracy of Pathfinder is 2.7 mm. Further calibrations of the robot are expected to improve this value significantly.

A linear time algorithm for computing exact Euclidean distance transforms of binary images in arbitrary dimensions

A sequential algorithm is presented for computing the exact Euclidean distance transform (DT) of a k-dimensional binary image in time linear in the total number of voxels N. The algorithm, which is based on dimensionality reduction and partial Voronoi diagram construction, can be used for computing the DT for a wide class of distance functions, including the L/sub p/ and chamfer metrics. At each dimension level, the DT is computed by constructing the intersection of the Voronoi diagram whose sites are the feature voxels with each row of the image. This construction is performed efficiently by using the DT in the next lower dimension. The correctness and linear time complexity are demonstrated analytically and verified experimentally. The algorithm may be of practical value since it is relatively simple and easy to implement and it is relatively fast (not only does it run in O(N) time but the time constant is small). A simple modification of the algorithm computes the weighted Euclidean DT, which is useful for images with anisotropic voxel dimensions. A parallel version of the algorithm runs in O(N/p) time with p processors.

Very High-Resolution Morphometry Using Mass-Preserving Deformations and HAMMER Elastic Registration

This article presents a very high-resolution voxel-based morphometric method, by using a mass-preserving deformation mechanism and a fully automated spatial normalization approach, referred to as HAMMER. By using a hierarchical attribute-based deformation strategy, HAMMER partly overcomes limitations of several existing spatial normalization methods, and it achieves a level of accuracy that makes possible morphometric measurements of spatial specificity close to the voxel dimensions. The proposed method is validated by a series of experiments, with both simulated and real brain images.

High-resolution 3D MRI of mouse brain reveals small cerebral structures in vivo

This work demonstrates technical approaches to high-quality magnetic resonance imaging (MRI) of small structures of the mouse brain in vivo. It turns out that excellent soft-tissue contrast requires the reduction of partial volume effects by using 3D MRI at high (isotropic) resolution with linear voxel dimensions of about 100–150 μm. The long T 2* relaxation times at relatively low magnetic fields (2.35 T) offer the benefit of a small receiver bandwidth (increased signal-to-noise) at a moderate echo time which together with the small voxel size avoids visual susceptibility artifacts. For measuring times of 1–1.5 h both T 1-weighted (FLASH) and T 2-weighted (Fast Spin-Echo) 3D MRI acquisitions exhibit detailed anatomical insights in accordance with histological sections from a mouse brain atlas. Preliminary applications address the identification of neuroanatomical variations in different mouse strains and the use of Mn 2+ as a T 1 contrast agent for neuroaxonal tracing of fiber tracts within the mouse visual pathway.

Evaluation of software for registration of contrast-enhanced brain MR images in patients with glioblastoma multiforme.

We evaluated commercially available software that rapidly and automatically registers brain MR images on a clinical workstation, and we studied the accuracy of these registrations. Ten patients with a diagnosis of glioblastoma multiforme underwent contrast-enhanced inversion recovery prepared three-dimensional (3D) volumetric spoiled gradient-recalled acquisition in the steady state (SPGR) MR imaging (contiguous 1.5-mm slice thickness, 96-104 slices). After this imaging sequence, each patient was brought out of the head coil into a sitting position and then repositioned in the coil. The inversion recovery prepared 3D SPGR sequence was then repeated. A commercially available software program operating on a clinical workstation was used to automatically register the second inversion recovery prepared SPGR series to the first. The speed of registration was recorded. The accuracy of each registration was estimated by recording the coordinates of eight anatomic landmarks on the registered and reference series and by calculating the mean error among matching landmarks. In nine of 10 patients, the registration software produced a visually satisfactory registration. In one patient, a second registration was necessary to produce a satisfactory registration. The processing time for each iteration was 48.3 +/- 3.8 sec (mean +/- SD). The mean error in aligning matching anatomic landmarks ranged from 0.67 to 1.41 mm, with an overall mean of 1.18 mm. The largest error among matching landmarks was 2.3 mm. Commercially available registration software can automatically register 3D imaging volumes in less than 1 min. The mean error in registration was approximately equivalent to the dimensions of a single voxel.

Signal‐to‐Noise Ratio as a Function of Imaging Parameters

AbstractSignal‐to‐Noise Ratio as a Function of Imaging Parameters (Azim Celik, General Electric Company, Milwaukee, Wisconsin and Weili Lin, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina). The degree to which noise affects a measurement is generally characterized by the signal‐to‐noise ratio (SNR, as measured by the ratio of the voxel signal to the noise standard deviation). This unit describes the importance of SNR in describing image quality. SNR is the key parameter for determining the quality of any given imaging experiment. If the SNR is not high enough, it becomes impossible to differentiate tissues from one another or the background. The dependence of SNR on imaging parameters such as the number of repetitions, the number of k‐space samples (Nx, Ny, and Nz), the readout bandwidth, and voxel dimensions (Dx, Dy, and Dz) is explained in detail.

Rapid sub-diffraction-limit laser micro/nanoprocessing in a threshold material system

We report on sub-diffraction-limit (SDL) micro/nanofabrication via two-photon-absorption (TPA) photopolymerization. SDL spatial resolution was found possible in a threshold system, where materials responded to light excitation with a pronounced threshold behavior. The diffraction limit in this case became just a measure of focal spot size, but did not put any actual restraint to voxel dimensions. Experimentally, lateral spatial resolution down to 120 nm was realized by using high numerical-aperture optics. In addition, we proposed a profile scanning method, by which the fabrication efficiency was significantly increased. The TPA processing time, in the example given here, was reduced by 90%.

Implementation and application of a brain template for multiple volumes of interest

We present a region template and a protocol for transforming that template to define anatomical volumes of interest (VOIs) in the human brain without operator intervention, based on software contained in the SPM99 package (Statistical Parametric Mapping, Wellcome Department of Cognitive Neurology, London, UK). We used an MRI of a reference brain to create an anatomical template of 41 VOIs, covering the entire brain, that can be spatially transformed to fit individual brain scans. Modified software allows for the reslicing and adaptation of the transformed template to any type of coregistered functional data. Individually defined VOIs can be added. We present an assessment of the necessary spatial transformations and compare results obtained for scans acquired in two different orientations. To evaluate the spatial transformations, 11 landmarks distributed throughout the brain were chosen. Euclidean distances between repeat samples at each landmark were averaged across all landmarks to give a mean difference of 1.3 plus minus 1.0 mm. Average Euclidean distances between landmarks (MRI:transformed template) were 8.1 plus minus 3.7 mm in anterior-posterior commissure (ACPC) and 7.6 plus minus 3.7 mm in temporal lobe (TL) orientation. In this study, we use [(11)C]-flumazenil-(FMZ-)PET as an example for the application of the region template. Thirty-four healthy volunteers were scanned, 21 in standard ACPC orientation, 13 in TL orientation. All had high resolution MRI and FMZ-PET. The average coefficient of variation (CV) of FMZ binding for cortical regions was 0.15, comparable with CVs from manually defined VOIs. FMZ binding was significantly different in 6/19 anatomical areas in the control groups obtained in the different orientations, probably due to anisotropic voxel dimensions. This new template allows for the reliable and fast definition of multiple VOIs. It can be used for different imaging modalities and in different orientations. It is necessary that imaging data for groups compared are acquired in the same orientation.

Mutual information-based rigid and nonrigid registration of ultrasound volumes

We investigated the registration of ultrasound volumes based on the mutual information measure, a technique originally applied to multimodality registration of brain images. A prerequisite for successful registration is a smooth, quasi-convex mutual information surface with an unambiguous maximum. We discuss the necessary preprocessing to create such a surface for ultrasound volumes. Abdominal and thoracic organs imaged with ultrasound typically move relative to the exterior of the body and are deformable. Consequently, four specific instances of image registration involving progressively generalized transformations were studied: rigid-body, rigid-body + uniform scaling, rigid-body + nonuniform scaling, and affine. Registration was applied to clinically acquired volumetric images. The accuracy was comparable with the voxel dimension for all transformation modes, although it degraded as the transformation grew more complex. Likewise, the capture range became narrower with the complexity of transformation. As the use of real-time three-dimensional ultrasound becomes more prevalent, the method we present should work well for a variety of applications examining serial anatomic and physiologic changes. Developers of these clinical applications would match the deformation model of their problem to one of the four transformation models presented here.

SNR versus resolution in 3D 1H MRS of the human brain at high magnetic fields.

It is commonly accepted that the signal-to-noise ratio (SNR = peak-signal/RMS-noise) per-unit-time of proton MR spectroscopy (1H-MRS) is linearly proportional to the voxel volume. Consequently, with a headcoil and 30-min acquisition, 1 cm3 is considered the SNR-limited spatial resolution barrier in the human brain. However, since local linewidths, Delta(upsilon*) = (piT2*)(-1), at high magnetic fields (B0), are dominated by regional inhomogeneities (DeltaB0), i.e., T2* << T2, reducing the voxel dimensions may increase T2*. This could compensate, in part, for signal loss with volume decrease. It is shown that for two cubic voxels of sides l1 and l2, l1 > l2, as the volume decreases by (l1/l2)3, their SNR ratio is reduced by only (l1/l2)2 due to a commensurate T2* increase of l1/l2. This is demonstrated in a phantom and the brains of volunteers, with 3D 1H-MRS in a headcoil at 4 T. It is shown that while the cubic voxels' dimensions were all halved, reducing their volume eightfold, their metabolites' SNR decreased only fourfold, due to their Delta(upsilon*s') twofold decrease. In other words, both spatial and spectral resolutions were doubled at a significantly, x2, smaller-than-expected SNR loss. This advantage was exploited to produce quality high spatial resolution, 0.75 x 0.75 x 0.75 cm3, metabolic maps in a 27-min acquisition.

Development of a 1.0 T MR microscope using a Nd-Fe-B permanent magnet

A compact 1.0 T MR microscope was developed using a permanent magnet made of high performance Nd-Fe-B magnetic material and a compact MRI console (54 cm (W) × 77 cm (H) × 60 cm (D), 80 kg weight). Since the magnetic field of the permanent magnet had a large temperature coefficient (−1200 ppm/deg), an internal NMR locking technique was developed for the imaging sequences. The performance of the system was evaluated using several biological specimens. As a result, good SNR 3D images at (50 μm)3–(200 μm)3 voxel dimensions were obtained in practical imaging times (0.5–7.5 hours). Thus we have concluded that the permanent-magnet compact MR microscope has great promise as a research or analytical tool.

High‐resolution fMRI using multislice partial k‐space GR‐EPI with cubic voxels

The premises of this work are: 1) the limit of spatial resolution in fMRI is determined by anatomy of the microcirculation; 2) because of cortical gray matter tortuosity, fMRI experiments should (in principle) be carried out using cubic voxels; and 3) the noise in fMRI experiments is dominated by low-frequency BOLD fluctuations that are a consequence of spontaneous neuronal events and are pixel-wise dependent. A new model is proposed for fMRI contrast which predicts that the contrast-to-noise ratio (CNR) tends to be independent of voxel dimensions (in the absence of partial voluming of activated tissue), TE, and scanner bandwidth. These predictions have been tested at 3 T, and results support the model. Scatter plots of fMRI signal intensities and low-frequency fluctuations for activated pixels in a finger-tapping paradigm demonstrated a linear relationship between signal and noise that was independent of TE. The R(2) value was about 0.9 across eight subjects studied. The CNR tended to be constant across pixels within a subject but varied across subjects: CNR = 3.2 +/- 1.0. fMRI statistics at 20- and 40-ms TE values were indistinguishable, and TE values as short as 10 ms were used successfully. Robust fMRI data were obtained across all subjects using 1 x 1 x 1 mm(3) cubic voxels with 10 contiguous slices, although 1.5 x 1.5 x 1.5 mm(3) was found to be optimum. Magn Reson Med 46:114-125, 2001.

Analysis of partial volume effects in diffusion-tensor MRI.

The diffusion tensor is currently the accepted model of diffusion in biological tissues. The measured diffusion behavior may be more complex when two or more distinct tissues with different diffusion tensors occupy the same voxel. In this study, a partial volume model of MRI signal behavior for two diffusion-tensor compartments is presented. Simulations using this model demonstrate that the conventional single diffusion tensor model could lead to highly variable and inaccurate measurements of diffusion behavior. The differences between the single and two-tensor models depend on the orientations, fractions, and exchange between the two diffusion tensor compartments, as well as the diffusion-tensor encoding technique and diffusion-weighting that is used in the measurements. The current single compartment model's inaccuracies could cause diffusion-based characterization of cerebral ischemia and white matter connectivity to be incorrect. A diffusion-tensor MRI imaging experiment on a normal human brain revealed significant partial volume effects between oblique white matter regions when using very large voxels and large diffusion-weighting (b approximately 2.69 x 10(3) sec/mm(2)). However, the apparent partial volume effects in white matter decreased significantly when smaller voxel dimensions were used. For diffusion tensor studies obtained using typical diffusion-weighting values (b approximately 1 x 10(3) sec/mm(2)) partial volume effects are much more difficult to detect and resolve. More accurate measurements of multiple diffusion compartments may lead to improved confidence in diffusion measurements for clinical applications.