Contrast-driven approach to intracranial segmentation using a combination of T2- and T1-weighted 3D MRI data sets

PURPOSE: To develop a strategy for structural brain imaging when using FSL software for segmentation and subsequent volumetry. MATERIALS AND METHODS: Three-dimensional (3D) structural MRI of 1-mm isotropic resolution was performed on a 3-Tesla clinical imaging system. Prescribed signal evolution of a multiple spin-echo (SE) sequence with variable refocusing flip angle for T2 weighting, and a modified driven equilibrium Fourier transform (MDEFT) sequence were used for T1 weighting. Postprocessing included rigid-body coregistration, brain extraction, and segmentation using the tools of the FSL 3.2 software package. RESULTS: T2 weighting provided reliable delineation of the subarachnoidal space, while T1 weighting provided better segmentation of gray matter (GM) and white matter (WM). The combination of T1-weighted (T1-w) and T2-w data allowed the identification of a T2-hypointense class of "nonbrain" (NB) representing larger vessels and structures of connective tissue, as well as partial volume of bone and air-filled cavities. CONCLUSION: Brain extraction on T2-w data and subsequent segmentation of the combined T1- and T2-w intensity distribution into four classes are recommended.     

Optimizing 3D EPI for rapid T1-weighted imaging

Purpose

To investigate the use of 3D EPI for rapid T₁-weighted brain imaging, focusing on the RF pulse’s influence on the contrast between gray and white matter.

Methods

An interleaved 3D EPI sequence use partial Fourier and CAIPIRINHA sampling was used to acquire T₁-weighted brain volumes with isotropic resolution, low echo times, and low geometric distortions. Five different RF pulses were evaluated in terms of fat suppression performance and gray–white matter contrast. Two binomial RF pulses were compared to a single rectangular (WE-rect) RF pulse exciting only water, and two new RF pulses developed in this work, where one was an extension of the WE-rect, and the other was an SLR pulse. The technique was demonstrated in three clinical cases, where brain tumor patients were imaged before and after gadolinium administration.

Results

A fat-suppressed 3D EPI sequence with a phase encoding bandwidth of around 100 Hz was found to exhibit a good trade-off between geometrical distortions and scan duration. Whole-brain T₁-weighted 3D EPI images with 1.2 mm isotropic voxel size could be acquired in 24 seconds. The WE-rect, its extension, and the SLR RF pulses resulted in reduced magnetization transfer effects and provided a 20% mean increase in gray–white matter contrast.

Conclusion

Using a high phase encoding bandwidth and RF pulses that reduce magnetization transfer effects, a fat-suppressed multi-shot 3D EPI sequence can be used to rapidly acquire isotropic T₁-weighted volumes.

     

Dynamic contrast-enhanced quantitative perfusion measurement of the brain using T1-weighted MRI at 3T

PURPOSE: To develop a method for the measurement of brain perfusion based on dynamic contrast-enhanced T(1)-weighted MR imaging. MATERIALS AND METHODS: Dynamic imaging of the first pass of a bolus of a paramagnetic contrast agent was performed using a 3T whole-body magnet and a T(1)-weighted fast field echo sequence. The input function was obtained from the internal carotid artery. An initial T(1) measurement was performed in order to convert the MR signal to concentration of the contrast agent. Pixelwise and region of interest (ROI)-based calculation of cerebral perfusion (CBF) was performed using Tikhonov's procedure of deconvolution. Seven patients with acute optic neuritis and two patients with acute stroke were investigated. RESULTS: The mean perfusion value for ROIs in gray matter was 62 mL/100g/min and 21 mL/100g/min in white matter in patients with acute optic neuritis. The perfusion inside the infarct core was 9 mL/100g/min in one of the stroke patients. The other stroke patient had postischemic hyperperfusion and CBF was 140 mL/100g/min. CONCLUSION: Absolute values of brain perfusion can be obtained using dynamic contrast-enhanced MRI. These values correspond to expected values from established PET methods. Furthermore, at 3T pixelwise calculation can be performed, allowing construction of CBF maps.     

3D magnetization prepared elliptical centric fast gradient echo imaging.

3D magnetization-prepared fast gradient echo MR sequences, such as MP-RAGE and IR-SPGR, provide good spatial resolution and gray-white contrast. The efficiency and image quality of these techniques can be further improved with an interleaved, recessed elliptical centric view order. It is shown that this novel acquisition strategy, along with skipping the acquisition of views in k-space corners can provide images with higher signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), while reducing artifact level and scan time compared to standard MP-RAGE.     

Magnetic resonance imaging of the cervical spine: comparison of 2D T2-weighted turbo spin echo, 2D T2*weighted gradient-recalled echo and 3D T2-weighted variable flip-angle turbo spin echo sequences

To compare an isotropic three-dimensional (3D) high-resolution T2-weighted (w) MR sequence and its reformations with conventional sequences for imaging of the cervical spine. Fifteen volunteers were examined at 1.5 T using sagittal and axial 3D T2-w, sagittal and axial 2D T2w, and axial 2D T2*w MR sequences. Axial reformations of the sagittal 3D dataset were generated (3D MPR T2w). Signal-to-noise and image homogeneity were evaluated in a phantom and in vivo. Visibility of ten anatomical structures of the cervical spine was evaluated. Artifacts were assessed. For statistical analysis, Cohen’s kappa, Wilcoxon matched pairs, and t-testing were utilized. There were no significant differences in homogeneity between the sequences. Sagittal 3D T2w enabled better delineation of nerve roots, neural foramina, and intraforaminal structures compared to sagittal 2D T2w. Axial 3D T2w and axial 3D MPR T2w resulted in superior visibility of most anatomical structures compared to axial 2D T2w and comparable results to 2D T2*w concerning the spinal cord, nerve roots, intraforaminal structures, and fat. Artifacts were most pronounced in axial 2D T2w and axial 3D T2w. Acquisition of a 3D T2w data set is feasible in the cervical spine with superior delineation of anatomical structures compared to 2D sequences.     

Optimization of blood-myocardial contrast in 3D true FISP cardiac imaging at 1.5 T.

Three-dimensional (3D) steady-state free precession (SSFP) MRI sequences are often applied to visualize both intra- and extracardiac pathologies. In the present study the contrast behavior of 3D true fast imaging with steady precession (True-FISP) sequences for cardiac imaging was optimized in numerical simulations and compared with measurements obtained in eight healthy volunteers on a 1.5 T whole-body scanner. Two SS preparation schemes in combination with and without a T(2) preparation were assessed to improve contrast between blood and myocardium using a navigator-gated and ECG-triggered 3D True-FISP sequence. Numerical simulations and experimental studies in volunteers showed that an SS preparation using a constant flip angle (CFA) is preferable to a linear flip angle (LFA) preparation in terms of contrast between blood and myocardium. The optimized 3D True-FISP sequence provides a reliable, accurate, and time-efficient means of obtaining a morphological cardiac diagnosis.     

Spatially encoded phase‐contrast MRI—3D MRI movies of 1D and 2D structures at millisecond resolution

This work demonstrates that the principles underlying phase‐contrast MRI may be used to encode spatial rather than flow information along a perpendicular dimension, if this dimension contains an MRI‐visible object at only one spatial location. In particular, the situation applies to 3D mapping of curved 2D structures which requires only two projection images with different spatial phase‐encoding gradients. These phase‐contrast gradients define the field of view and mean spin‐density positions of the object in the perpendicular dimension by respective phase differences. When combined with highly undersampled radial fast low angle shot (FLASH) and image reconstruction by regularized nonlinear inversion, spatial phase‐contrast MRI allows for dynamic 3D mapping of 2D structures in real time. First examples include 3D MRI movies of the acting human hand at a temporal resolution of 50 ms. With an even simpler technique, 3D maps of curved 1D structures may be obtained from only three acquisitions of a frequency‐encoded MRI signal with two perpendicular phase encodings. Here, 3D MRI movies of a rapidly rotating banana were obtained at 5 ms resolution or 200 frames per second. In conclusion, spatial phase‐contrast 3D MRI of 2D or 1D structures is respective two or four orders of magnitude faster than conventional 3D MRI. Magn Reson Med, 2011. © 2011 Wiley‐Liss, Inc.     

T1 measurement of flowing blood and arterial input function determination for quantitative 3D T1-weighted DCE-MRI

PURPOSE: To propose a simple, accurate method for measuring T(1) in flowing blood and the arterial input function (AIF), and to evaluate the impact on dynamic contrast-enhanced MRI (DCE-MRI) quantification of pharmacokinetic parameters. MATERIALS AND METHODS: A total of 10 rabbits were scanned at 1.5 Tesla and administered a bolus of Gadomer. Preinjection T(1) and AIF measurements were acquired in the iliac arteries using a rapid three-dimensional (3D) spoiled gradient recalled echo (SPGR) approach. Correction was made for imperfect B(1) fields, in-flow, and partial volume effects. DCE-MRI parameters blood volume (v(b)) and endothelial transfer constant (K(trans)) in resting skeletal muscle were estimated from pharmacokinetic analysis using individually measured AIFs. Literature comparisons were made to assess accuracy. RESULTS: Blood T(1) was more accurate and precise after correction for B(1) and partial volume errors (1267 +/- 72 msec). Measured AIFs followed reported biexponential decay characteristics for Gadomer clearance in rabbits. Parameters v(b) (2.47 +/- 0.65%) and K(trans) (3.6 +/- 1.0 x 10(-3) minute(-1)) derived from AIFs based on corrected blood T(1)s were more reproducible and in better agreement with literature values. CONCLUSION: The proposed method enables accurate in vivo blood T(1) and AIF measurements and can be easily implemented in a range of DCE-MRI applications to improve both the accuracy and reproducibility of pharmacokinetic parameters.     

Assessment of relative brain iron concentrations using T2-weighted and T2*-weighted MRI at 3 Tesla

In this paper a new method is presented for the relative assessment of brain iron concentrations based on the evaluation of T₂ and T₂* -weighted images. A multiecho sequence is employed for rapid measurement of T₂ and T₂*, enabling calculation of the line broadening effect ( T₂′). Several groups have failed to show a correlation between T₂ and brain iron content. However, quantification of T₂′, and the associated relaxation rate R₂′, may provide a more specific relative measure of brain iron concentration. This may find application in the study of brain diseases, which cause associated changes in brain iron levels. A new method of field inhomogeneity correction is presented that allows the separation of global and local field inhomogeneities, leading to more accurate T₂* measurements and hence, T₂′ values. The combination of T₂*, and T₂-weighted MRI methods enables the differentiation of Parlkinson's disease patients from normal age-matched controls based on differences in iron content within the substantia nigra.     

Optimized T1-weighted contrast for single-slab 3D turbo spin-echo imaging with long echo trains: application to whole-brain imaging.

T(1)-weighted contrast is conventionally obtained using multislice two-dimensional (2D) spin-echo (SE) imaging. Achieving isotropic, high spatial resolution is problematic with conventional methods due to a long acquisition time, imperfect slice profiles, or high-energy deposition. Single-slab 3D SE imaging was recently developed employing long echo trains with variable low flip angles to address these problems. However, long echo trains may yield suboptimal T(1)-weighted contrast, since T(2) weighting of the signals tends to develop along the echo train. Image blurring may also occur if high spatial frequency signals are acquired with low signal intensity. The purpose of this work was to develop an optimized T(1)-weighted version of single-slab 3D SE imaging with long echo trains. Refocusing flip angles were calculated based on a tissue-specific prescribed signal evolution. Spatially nonselective excitation was used, followed by half-Fourier acquisition in the in-plane phase encoding (PE) direction. Restore radio frequency (RF) pulses were applied at the end of the echo train to optimize T(1)-weighted contrast. Imaging parameters were optimized by using Bloch equation simulation, and imaging studies of healthy subjects were performed to investigate the feasibility of whole-brain imaging with isotropic, high spatial resolution. The proposed technique permitted highly-efficient T(1)-weighted 3D SE imaging of the brain.     

T2rho-weighted contrast in MR images of the human brain.

In this work, the feasibility of using T2rho weighting as an MR contrast mechanism is evaluated. Axial images of a human brain were acquired using a single-slice spin-lock T2rho-weighted pulse sequence and compared to analogous T2-weighted images of the same slice. The contrast between white matter and gray matter in T2rho-weighted images was approximately 40% greater than that from T2-weighted data. These preliminary data suggest that the novel contrast mechanism of T2rho can be used to yield high-contrast T2-like images.     

3D MDEFT imaging of the human brain at 4.7 T with reduced sensitivity to radiofrequency inhomogeneity.

A modification to the 3D modified driven equilibrium Fourier transform (MDEFT) imaging technique is proposed that reduces its sensitivity to RF inhomogeneity. This is especially important at high field strengths where RF focusing effects exacerbate B(1) inhomogeneity, causing significant signal nonuniformity in the images. The adiabatic inversion pulse used during the preparation period of the MDEFT sequence is replaced by a hard (nonadiabatic) pulse with a nominal flip angle of 130 degrees. The spatial inhomogeneity of the hard pulse preparation compensates for the inhomogeneity of the excitation pulses. Uniform signal intensity is obtained for a wide range of B(1) amplitudes and the high CNR characteristic of MDEFT is retained. The new approach was validated by numerical simulations and successfully applied to human brain imaging at 4.7 T, resulting in high-quality T(1)-weighted images of the whole human brain at high field strength with uniform signal intensity and contrast, despite the presence of significant RF inhomogeneity.     

Embedded MR fluoroscopy: high temporal resolution real-time imaging during high spatial resolution 3D MRA acquisition.

A method termed "embedded fluoroscopy" for simultaneously acquiring a real-time sequence of 2D images during acquisition of a 3D image is presented. The 2D images are formed by periodically sampling the central phase encodes of the slab-select direction during the 3D acquisition. The tradeoffs in spatial and temporal resolution are quantified by two parameters: the "redundancy" (R), the fraction of the 3D acquisition sampled more than once; and the "effective temporal resolution" (T), the time between temporal updates of the central views. The method is applied to contrast-enhanced MR angiography (CE-MRA). The contrast bolus dynamics are portrayed in real time in the 2D image sequence while a high-resolution 3D image is being acquired. The capability of the 2D acquisition to measure contrast enhancement with only a 5% degradation of the spatial resolution of the 3D CE-MR angiogram is shown theoretically. The method is tested clinically in 15 CE-MRA patient studies of the carotid and renal arteries.     

Magnetic resonance imaging of the brain with gadopentetate dimeglumine-DTPA: Comparison of T1-weighted spin-echo and 3D gradient-echo sequences

Short TR, short TE, high resolution, 3D gradient-recalled echo (GRE) imaging was evaluated for lesion detection in the brain. High resolution 3D GRE data acquisition was used to reduce partial volume effects and flow artifacts, to better visualize smaller structures, to minimize signal losses caused by field inhomogeneities, and to allow better image reformatting. Spin-echo (SE) and 3D GRE approaches were compared for lesion detection after the administration of an MR contrast agent, gadopentetate dimeglumine. Preliminary clinical studies demonstrated that the signal-to-noise ratio (SNR) in each slice of the GRE scan was worse than that of the SE scan because of the much thicker slices acquired with the SE technique. However, by averaging two adjacent 3D slices, the SNR of the two methods was essentially equivalent. In the averaged GRE slices, large lesions were seen just as well as in the SE images. More importantly, small lesions were better visualized in the thin 3D GRE images than in the thick SE images for the lesions studied in this work and the protocols used. These observations were confirmed by theoretical simulations.     

Efficient method for calculating kinetic parameters using T1-weighted dynamic contrast-enhanced magnetic resonance imaging.

It has become increasingly important to quantitatively estimate tissue physiological parameters such as perfusion, capillary permeability, and the volume of extravascular-extracellular space (EES) using T(1)-weighted dynamic contrast-enhanced MRI (DCE-MRI). A linear equation was derived by integrating the differential equation describing the kinetic behavior of contrast agent (CA) in tissue, from which K(1) (rate constant for the transfer of CA from plasma to EES), k(2) (rate constant for the transfer from EES to plasma), and V(p) (plasma volume) can be easily obtained by the linear least-squares (LLSQ) method. The usefulness of this method was investigated by means of computer simulations, in comparison with the nonlinear least-squares (NLSQ) method. The new method calculated the above parameters faster than the NLSQ method by a factor of approximately 6, and estimated them more accurately than the NLSQ method at a signal-to-noise ratio (SNR) of < approximately 10. This method will be useful for generating functional images of K(1), k(2), and V(p) from DCE-MRI data.