Introduction MRI provides a noninvasive method for quantifying trabecular bone architecture for the purpose of assessing fracture risk in subjects with osteopenia (1, 2). However, as trabecular thickness is on the order of the achievable voxel size in vivo (100-150μm), more sophisticated processing and analysis techniques are required for accurate extraction of structural parameters. Linear interpolation is commonly applied to increase the apparent resolution of digital images. In one dimension, for example, the bone volume fraction (BVF) at a spatial location between the centers of two adjacent voxels would be computed as the average of the two voxels. Therefore, additional values calculated in this manner will never increase beyond the original values and, thus, contradict the notion that smaller voxels are more likely to contain larger fractions of bone. The subvoxel processing technique presented here reduces partial volume blurring effects, building on a previously developed method for estimating voxel BVF’s from images acquired in the limited spatial resolution regime (3). Theory Subvoxel processing rests on two assumptions: (1) smaller voxels are more likely to have high BVF and (2) bone is generally in close proximity to more bone. Each voxel is partitioned into eight subvoxels. The algorithm enforces strict conservation of bone mass, i.e. the total BVF in the original voxel is divided among the subvoxels. The precise amount allotted to a subvoxel is determined by the amount and location of bone outside the voxel but adjacent to the subvoxel. Thus, bone tends to be sequestered in the area of the voxel which is closest to other bone. The principle is illustrated in 2D with a 3×3 array of voxels in which the central voxel (BVF=0.2) has been partitioned into four subvoxels (Fig. 1). The weight for subvoxel a is the sum of the three adjacent voxel BVF’s (i.e., 0.3+0.4+0.6=1.3). Similarly, the weights for subvoxels b and c are 0+0.5+0.6=1.1 and zero, respectively. Subvoxels are assigned a nonzero weight only if bone is “attracted” to both outer sides by voxels containing bone. For example, subvoxel b has two voxels with BVF=0.6 and BVF=0.5 attracting bone to the right. The voxel with BVF=0.5 also attracts bone towards the bottom. In contrast, subvoxel d has no voxels attracting bone from the left and is thus assigned zero weight. Therefore, 1.3/(1.3+1.1)=0.54 and 1.1/(1.3+1.1)=0.46 of the bone is allotted to subvoxels a and b, respectively, yielding 0.54•0.2•4=0.43 and 0.46•0.2•4=0.37 (Fig. 1B). The factor of four is necessary since the volume of the subvoxel is four times smaller than that of the voxel. If voxel BVF is high, a subvoxel may be assigned an unrealistic BVF>1. In such a case, the subvoxel is assigned a BVF of 1 and the excess bone is assigned to the other subvoxels. The algorithm is easily applicable to all three spatial dimensions. Materials and Methods Subvoxel processing of an in vivo 3D MR image of the distal radius of a 60-year-old woman (Fig. 2) was compared to trilinear interpolation. The image was acquired at 1.5T using FLASE (4) at a voxel size of 137×137×350μm. A voxel BVF map was generated, after applying iterative deconvolution to estimate the noiseless histogram of voxel intensities (3). A cylindrical VOI (31×31×22 voxels) was arbitrarily chosen (Fig. 1B). Subvoxel processing and trilinear interpolation were applied to increase the apparent resolution in all three dimensions.
The strength of trabecular bone and its resistance to fracture traditionally have been associated with apparent density. This paradigm assumes that neither the ultrastructural nor microstructural make-up of the bone is altered during aging and osteoporosis. During the past decade there has been growing evidence from both laboratory and clinical studies against this view. Recent advances in noninvasive imaging technology, notably micro-magnetic resonance imaging (μ MRI) and computed tomography, offer an opportunity to test the hypothesis that architecture is an independent contributor to bone strength. MRI appears to be ideally suited for this task because bone marrow has uniform high signal intensity while bone appears with background intensity, thus yielding a binary system tomographic system. However, in vivo trabecular bone imaging is hampered by the limited signal-to-noise ratio that precludes voxel sizes much smaller than trabecular thickness, which would be required to yield a bimodal intensity histogram for segmentation of the image into bone and marrow. The resulting partial volume blurring leads to fuzzy boundaries. Successful structure analysis thus demands more elaborate processing strategies. This article reviews new approaches conceived in the authors' laboratory toward acquisition, processing, and structural analysis of trabecular bone images in the limited spatial resolution regimen of in vivo μ MRI. These methods are shown to provide detailed insight into the three-dimensional trabecular network topology and scale at the distal radius or distal tibia that typically serve as surrogate sites. The μ MRI–derived structural parameters are shown to be associated with the bone's biomechanical properties and fracture resistance. Further, the technology has advanced to a stage permitting serial studies in laboratory animals and humans as a means to evaluate the effects of treatment. The method currently is confined to peripheral skeletal sites, and its extension to typical fracture sites such as the proximal femur hinges on further advances in detection sensitivity.
Purpose To compare calf skeletal muscle perfusion measured with pulsed arterial spin labeling (PASL) and pseudo‐continuous arterial spin labeling (pCASL) methods, and to assess the variability of pCASL labeling efficiency in the popliteal artery throughout an ischemia‐reperfusion paradigm. Materials and Methods At 3T, relative pCASL labeling efficiency was experimentally assessed in five subjects by measuring the signal intensity of blood in the popliteal artery just distal to the labeling plane immediately following pCASL labeling or control preparation pulses, or without any preparation pulses throughout separate ischemia‐reperfusion paradigms. The relative label and control efficiencies were determined during baseline, hyperemia, and recovery. In a separate cohort of 10 subjects, pCASL and PASL sequences were used to measure reactive hyperemia perfusion dynamics. Results Calculated pCASL labeling and control efficiencies did not differ significantly between baseline and hyperemia or between hyperemia and recovery periods. Relative to the average baseline, pCASL label efficiency was 2 ± 9% lower during hyperemia. Perfusion dynamics measured with pCASL and PASL did not differ significantly ( P > 0.05). Average leg muscle peak perfusion was 47 ± 20 mL/min/100g or 50 ± 12 mL/min/100g, and time to peak perfusion was 25 ± 3 seconds and 25 ± 7 seconds from pCASL and PASL data, respectively. Differences of further metrics parameterizing the perfusion time course were not significant between pCASL and PASL measurements ( P > 0.05). Conclusion No change in pCASL labeling efficiency was detected despite the almost 10‐fold increase in average blood flow velocity in the popliteal artery. pCASL and PASL provide precise and consistent measurement of skeletal muscle reactive hyperemia perfusion dynamics. J. MAGN. RESON. IMAGING 2016;44:929–939.
OBJECTIVES/SPECIFIC AIMS: Computed tomography (CT) enables 3-dimensional (3D) visualization of cortical bone structures with high spatial resolution, and thus has been the gold-standard method for evaluation and diagnosis of craniofacial skeletal pathologies. However, ionizing radiation and, in particular, repeated scanning for presurgery and postsurgery assessments, is of concern when applied to infants and young children. Recent advances in solid-state MRI allow the capture of the short-T2 signals in cortical bone while suppressing the signal from soft-tissue protons having T2 relaxation time 1–2 orders of magnitude longer (50–100 ms). One approach, a dual-radiofrequency (RF) pulse and ultrashort echo time (UTE) imaging based method, exploits different sensitivities of bone and soft tissue to different RF pulse widths and TEs. This study aims to demonstrate the feasibility of producing 3D renderings of the human skull and visualization of cranial sutures using the bone-selective MRI technique in comparison to CT. METHODS/STUDY POPULATION: Imaging technique: Two RF pulses differing in duration and amplitude are alternately applied in successive repetition time (TR) along the pulse train. Within each TR, 2 echoes are acquired. Acquisition of the first echo starts at the ramp-up of the encoding gradient (TE1), allowing for capture of signals with very short lifetimes (bone), while that of the second starts after a longer delay (TE2). In total, 4 echoes are obtained: ECHO11 (RF1TE1), ECHO12 (RF1TE2), ECHO21 (RF2TE1), and ECHO22 (RF2TE2). During reconstruction, ECHO11 is combined with ECHO21 and ECHO12 is combined with ECHO22, resulting in 2 images. The subtraction of these 2 images yields an enhanced bone contrast. Data acquisition/processing: The pulse sequence described above was applied for MR imaging of a human cadaveric skull and 2 adult human subjects in vivo, at 3T field strength (Siemens Prisma, Erlangen, Germany). Imaging parameters: TR/TE1/TE2=7/0.06/2.46 ms, RF1/RF2 durations=40/520 μs, flip angle=12°, matrix size=2563, field of view=2803 mm 3 , voxel size=1.1 mm isotropic, number of radial spokes=25,000, and scan time=6 minutes. Segmentation of bone voxels was performed using ITK-SNAP in a semi-automatic fashion, leading to 3D renderings of the skull. For comparison, a CT scan was also performed in the human cadaveric skull with 1 mm isotropic resolution. Validation: The biometric accuracy was assessed by measuring eight anatomic distances: (1) Maximum craniocaudal aperture of the right orbit. (2) Maximum craniocaudal aperture of the left orbit. (3) Maximum height of the mandible from chin point in the midline. (4) Maximum cranial length (5) Maximum cranial width. (6) Maximum height of piriform aperture. (7) Distance between lateral most aspect of mandibular condyles. (8) Distance between lateral most aspect of posterior hard palate in both CT- and MRI-based 3D renderings of the human cadaveric skull using Mimics software (Materialise ® , Ghent, Belgium). These distances were compared with those directly measured on the cadaveric skull. RESULTS/ANTICIPATED RESULTS: Compares CT with the proposed MRI method on cadaveric human skull images, along with corresponding 3D renderings. Compared with CT, the 3D rendered images maintain most features over the entire head (e.g., zygomatic arch), except for appearance of some artifacts in the mandibular region. In vivo head images in 2 adult subjects: axial magnitude images and 3D rendering. In the axial images, bone voxels as well as the inner table of the cranium are clearly visualized, and cranial and spinal bone structures are well depicted in the 3D renderings. Some voxels were erroneously included or excluded in the renderings. The mean difference in measurements of the 8 anatomic distances was 6, 4, and 2 mm when comparing MRI Versus CT, MRI Versus in situ, and CT Versus in situ, respectively. DISCUSSION/SIGNIFICANCE OF IMPACT: Bone proton magnetization exhibits a substantial level of signal decay during the relatively long duration of RF2 due to its very short T2 relaxation time. In contrast, soft-tissue retains nearly the same level of signal intensities over all echoes. Thus, subtraction of ECHO22 from ECHO11, when compared with the difference between ECHO11 and ECHO12, enhances bone contrast from soft tissue. The proposed, dual-RF dual-echo 3D UTE imaging technique produces isotropic high-resolution bone-specified images in the whole head within a clinically feasible imaging time (6 min), leading to clear visualization of craniofacial skeletal structures. These are key components necessary for translation to the clinical setting. Optimization of postprocessing for more realistic 3D renderings and thus accurate anatomic measurements is currently being implemented. The proposed method’s potential as a nonradiative alternative to CT will then be thoroughly evaluated in pediatric patients.
PURPOSE: To develop and apply a method for the derivation of cancellous bone architectural parameters from in vivo magnetic resonance (MR) images of the distal radius and to evaluate these parameters as predictors of vertebral fracture status in osteopenia. MATERIALS AND METHODS: MR images (137 x 137 x 500-micron3 voxel size) were acquired with a three-dimensional partial flip-angle spin-echo pulse sequence in the distal radius of 36 women. Subjects were classified as healthy or with osteoporosis on the basis of vertebral deformity and bone mineral density (BMD). Images rated as of adequate quality in 20 subjects were processed with a method that is applicable in the limited spatial resolution regime. The method relies on histogram deconvolution to obviate binary segmentation. Cancellous bone structure was treated as a quasi-regular lattice and analyzed with spatial autocorrelation, yielding parameters that quantify intertrabecular spacing, contiguity, and a measure of longitudinal alignment called tubularity. RESULTS: Whereas neither BMD nor any of the structural parameters individually correlated significantly with vertebral deformity fraction, a simple function that involved tubularity and longitudinal spacing predicted deformity fraction well (r = .78, P < .005). CONCLUSION: Histomorphometric parameters characterizing cancellous bone in the distal radius can be derived from in vivo MR microimages and are predictive of vertebral deformity.
Six patients with pituitary abnormalities and three normal volunteers were evaluated by high field superconductive (1.0, 1.4, or 1.5 Tesla) magnetic resonance (MR) imaging, low field resistive (0.12 Tesla) MR imaging, and contrast-enhanced, high-resolution CT. Four macroadenomas, one microadenoma, and one empty sella were demonstrated. Their morphology and anatomic relationship to the visual pathway and the internal carotid and anterior cerebral arteries were best demonstrated by high field MR imaging. The low field resistive MR studies were least effective in showing the lesions.
Abstract A key function of sleep is to provide a regular period of reduced brain metabolism, which is critical for maintenance of healthy brain function. The purpose of this work was to quantify the sleep‐stage‐dependent changes in brain energetics in terms of cerebral metabolic rate of oxygen (CMRO 2 ) as a function of sleep stage using quantitative magnetic resonance imaging (MRI) with concurrent electroencephalography (EEG) during sleep in the scanner. Twenty‐two young and older subjects with regular sleep hygiene and Pittsburgh Sleep Quality Index (PSQI) in the normal range were recruited for the study. Cerebral blood flow (CBF) and venous oxygen saturation (SvO 2 ) were obtained simultaneously at 3 Tesla field strength and 2.7‐s temporal resolution during an 80‐min time series using OxFlow, an in‐house developed imaging sequence. The method yields whole‐brain CMRO 2 in absolute physiologic units via Fick's Principle. Nineteen subjects yielded evaluable data free of subject motion artifacts. Among these subjects, 10 achieved slow‐wave (N3) sleep, 16 achieved N2 sleep, and 19 achieved N1 sleep while undergoing the MRI protocol during scanning. Mean CMRO 2 was 98 ± 7(μmol min −1 )/100 g awake, declining progressively toward deepest sleep stage: 94 ± 10.8 (N1), 91 ± 11.4 (N2), and 76 ± 9.0 μmol min −1 /100 g (N3), with each level differing significantly from the wake state. The technology described is able to quantify cerebral oxygen metabolism in absolute physiologic units along with non‐REM sleep stage, indicating brain oxygen consumption to be closely associated with depth of sleep, with deeper sleep stages exhibiting progressively lower CMRO 2 levels.
