Automated spectral analysis III: Application to in Vivo proton MR Spectroscopy and spectroscopic imaging

An automated method for analysis of in vivo proton magnetic resonance (MR) spectra and reconstruction of metabolite distributions from MR spectroscopic imaging (MRSI) data is described. A parametric spectral model using acquisition specific, a priori information is combined with a wavelet-based, nonparametric characterization of baseline signals. For image reconstruction, the initial fit estimates were additionally modified according to a priori spatial constraints. The automated fitting procedure was applied to four different examples of MRS data obtained at 1.5 T and 4.1 T. For analysis of major metabolites at medium TE values, the method was shown to perform reliably even in the presence of large baseline signals and relatively poor signal-to-noise ratios typical of in vivo proton MRSI. identification of additional metabolites was also demonstrated for short TE data. Automated formation of metabolite images will greatly facilitate and expand the clinical applications of MR spectroscopic imaging.     

Assessment of 3D proton MR echo-planar spectroscopic imaging using automated spectral analysis.

For many clinical applications of proton MR spectroscopic imaging (MRSI) of the brain, diagnostic assessment is limited by insufficient coverage provided by single- or multislice acquisition methods as well as by the use of volume preselection methods. Additionally, traditional spectral analysis methods may limit the operator to detailed analysis of only a few selected brain regions. It is therefore highly desirable to use a fully 3D approach, combined with spectral analysis procedures that enable automated assessment of 3D metabolite distributions over the whole brain. In this study, a 3D echo-planar MRSI technique has been implemented without volume preselection to provide sufficient spatial resolution with maximum coverage of the brain. Using MRSI acquisitions in normal subjects at 1.5T and a fully automated spectral analysis procedure, an assessment of the resultant spectral quality and the extent of viable data in human brain was carried out. The analysis found that 69% of brain voxels were obtained with acceptable spectral quality at TE = 135 ms, and 52% at TE = 25 ms. Most of the rejected voxels were located near the sinuses or temporal bones and demonstrated poor B0 homogeneity and additional regions were affected by stronger lipid contamination at TE = 25 ms.     

Fast proton spectroscopic imaging using the sliced k-space method

The use of one-shot imaging methods for proton spectroscolpic imaging (¹H-SI) is examined. In particular the acquisition of K_x × K_y × N_t data points by means of N_t excitations, each acquiring a K_x, × K_y, k-space slice, is advocated. A number of strategies for realising this experiment, and combining it with water suppression and volume-selection are proposed. The practical implementation at 4.7 T for ¹H-Sl of the rat brain is described. Experimental results from a 32 × 32 spatial matrix with N_t = 64 are presented. Spectra obtained from volumes as low as 3.5 μl and within measuring times of as little as 3.8 min are shown. In these choline, creatine/phosphocreatine and N-acetylaspartate are all clearly visible.     

Fast proton spectroscopic imaging of human brain using multiple spin-echoes

We introduce a multi-echo multi-slice MR proton spectroscopic imaging method, which allows for a dramatic reduction of the measurement time by acquiring multiple spin-echoes within a single repetition time. In the multi-echo multi-slice experiment discussed in this paper, a threefold reduction in measurement time is obtained by sacrificing some spectral resolution. Signal-to-noise ratio and spatial resolution are preserved. Metabolite images of N-acetyl aspartate, and total choline + total creatine from multiple slices through the human brain are presented and compared with images obtained with a conventional single-echo multi-slice method.     

Fast proton spectroscopic imaging employing k-space weighting achieved by variable repetition times

A k-space weighted spectroscopic imaging (SI) method is presented that allows a reduction in the total data acquisition time by up to 55% compared with standard SI. The k-space weighting is achieved by varying the repetition time, thus realizing an inherent apodization that corresponds to a circularly symmetric generalized Hamming filter. The flip angle is varied with the repetition time to enhance the signal-to-noise ratio. These techniques were employed using a short echo time of 10 ms. In vivo measurements on healthy rat brain at 4.7 T were conducted, obtaining two-dimensional spectroscopic imaging data from a 25 × 25 circularly reduced k-space area in as little as 5 min. The signal-to-noise ratio is sufficiently high to detect J-coupled resonances such as myo-inositol or glutamate/glutamine, demonstrating the ability to combine short acquisition times with comprehensive metabolic information. The T₁ dependency of the apodization and the corresponding point spread function was evaluated by computer simulations. The achievable signal-to-noise ratio per unit time was compared with standard SI giving a parameter-dependent advantage of approximately 20% of the standard SI method.     

Automated processing for proton spectroscopic imaging using water reference deconvolution

Automated formation of MR spectroscopic images (MRSI) is necessary before routine application of these methods is possible for in vivo studies; however, this task is complicated by the presence of spatially dependent instrumental distortions and the complex nature of the MR spectrum. A data processing method is presented for completely automated formation of in vivo proton spectroscopic images, and applied for analysis of human brain metabolites. This procedure uses the water reference deconvolution method (G. A. Morris, J. Magn. Reson. 80,547(1988)) to correct for line shape distortions caused by instrumental and sample characteristics, followed by parametric spectral analysis. Results for automated image formation were found to compare favorably with operator dependent spectral integration methods. While the water reference deconvolution processing was found to provide good correction of spatially dependent resonance frequency shifts, it was found to be susceptible to errors for correction of line shape distortions. These occur due to differences between the water reference and the metabolite distributions.     

Double-echo multislice proton spectroscopic imaging using Hadamard slice encoding

Simultaneous multislice proton spectroscopic imaging (SI) is presented using a pulse sequence with multifrequency selective RF excitation and Hadamard encoding in the slice direction, and conventional Fourier phase encoding in the in-plane directions. Double-echo data acquisition is used to increase the spectral information of the experiment. Tests on a phantom demonstrate the quality of the slice selection. Results of in vivo measurements on the healthy rat brain show that spectra with a high signal-to-noise ratio can be acquired from four slices within 32 min. The measurements were performed at 4.7 T using a field of view of 32 × 32 mm², a slice thickness of 3 mm, and a voxel size of 12 μl. The proposed method is a useful alternative to sequential multislice SI and 3D SI. Furthermore, it is possible to combine sequential and simultaneous multi-slice SI.     

Evaluation of variable line-shape models and prior information in automated 1H spectroscopic imaging analysis.

Analysis of in vivo short TE 1H spectra is complicated by broad baseline signal contributions and resonance line-shape distortions. Although the assumptions of ideal metabolite resonance line-shapes and slowly varying baseline signals can be used to separate these signals, the presence of broad or asymmetric line-shapes can invalidate this model. More complex line-shape models are computationally expensive or difficult to constrain, particularly for the low signal-to-noise commonly found for in vivo MR spectroscopic imaging applications. In this study, two time-domain models for fitting variable spectral line-shapes are examined, one using B-splines and another using summed sinusoids. The methods were verified using both phantom and human data, and Monte Carlo simulations were used to evaluate variations in calculated metabolite amplitudes due to interactions between the baseline and line-shape estimations. Additional studies investigated the use of prior line-shape information, obtained from either a water MRSI measurement or calculations from B(0) maps, to determine parameter starting values or optimization constraints. Both line-shape models showed the ability to fit the variety of line-shapes present in both the phantom and human MRSI data, with similar or improved accuracy over a Gaussian line-shape model; however, this improvement resulted in only minor improvement for the high-SNR phantom data and moderate improvements in regions with asymmetry for the fitted in vivo metabolite images. The use of prior line-shape information was of most benefit when applied toward setting optimization constraints but was of limited benefit when used to define initial starting values.     

2D 1H spectroscopic imaging of the human brain at 4.1 T

A two-dimensional spectroscopic imaging sequence consisting of an inversion recovery pulse, a plane selective prefocused pulse, and a semiselective water suppression pulse has been used to create ¹H spectroscopic images of the human brain with nominal voxels of 0.5 cc. Due to the excellent lipid suppression provided by the inversion recovery pulse and subsequent delay, only planar volume selection is required enabling the entire brain within the slice to be imaged without contamination from extracerebral lipids in the brain voxels. The use of a semiselective refocusing pulse for water suppression permits any echo evolution time to be used, minimizing J-modulation and T₂ losses, while retaining full sensitivity in the lactate resonance. Using this sequence we have visualized the lactate elevation in the peri-infarct region about a 6-week-old stroke.     

Temporal and regional changes during focal ischemia in rat brain studied by proton spectroscopic imaging and quantitative diffusion NMR imaging

The early development of focal ischemia after permanent occlusion of the right middle cerebral artery (MCA) was studied in six rats using interleaved measurements by diffusion-weighted NMR imaging (DWI) of water and two variants of proton spectroscopic imaging (SI), multiecho SI (TE: 136, 272, 408 ms) and short TE SI (TE: 20 ms). Measurements on a 4.7-T NMR imaging system were performed between the control phase and approximately 6 h postocclusion. In the center of the ischemic lesion of all rats, the apparent diffusion coefficient (ADC) decreased rapidly to 84.4 ± 4.2% (mean ± SD) of the control values approximately 2 min postocclusion. Approximately 6 h postocclusion, the ADC was reduced to 67.1 ± 5.9%. In contrast, large differences between the animals were observed for the temporal increase of lactate (Lac) in the ipsilateral hemisphere. The maximum Lac signal was reached in four rats after 0.5-1.5 h, and in two rats was not reached even after 6 h postocclusion. Six h postocclusion, SI spectra measured at a TE of 136 ms revealed a decrease in the CH³ signal of N-acetylaspartate (NAA) to 67 ± 13% of the control values. Differences were observed between the spatial regions of decreased NAA and increased Lac. In the lesions, a T2 relaxation time of Lac of 292 ± 40 ms, considering a J-cou-pling constant of 6.9 Hz, was measured. Furthermore, a prolongation of the T₂ of the CH₃ signal of creatine/phosphocre-atine (Cr/PCr) was observed in the lesion, from 163 ± 22 ms during control to 211 ± 41 ms approximately 6 h postocclusion. The experiments proved that DWI and proton SI are valuable tools to provide complementary information on processes associated with brain infarcts.     

Fast proton spectroscopic imaging using steady-state free precession methods.

Various pulse sequences for fast proton spectroscopic imaging (SI) using the steady-state free precession (SSFP) condition are proposed. The sequences use either only the FID-like signal S(1), only the echo-like signal S(2), or both signals in separate but adjacent acquisition windows. As in SSFP imaging, S(1) and S(2) are separated by spoiler gradients. RF excitation is performed by slice-selective or chemical shift-selective pulses. The signals are detected in absence of a B(0) gradient. Spatial localization is achieved by phase-encoding gradients which are applied prior to and rewound after each signal acquisition. Measurements with 2D or 3D spatial resolution were performed at 4.7 T on phantoms and healthy rat brain in vivo allowing the detection of uncoupled and J-coupled spins. The main advantages of SSFP based SI are the short minimum total measurement time (T(min)) and the high signal-to-noise ratio per unit measurement time (SNR(t)). The methods are of particular interest at higher magnetic field strength B(0), as TR can be reduced with increasing B(0) leading to a reduced T(min) and an increased SNR(t). Drawbacks consist of the limited spectral resolution, particularly at lower B(0), and the dependence of the signal intensities on T(1) and T(2). Further improvements are discussed including optimized data processing and signal detection under oscillating B(0) gradients leading to a further reduction in T(min).     

Mapping of brain metabolite distributions by volumetric proton MR spectroscopic imaging (MRSI)

Distributions of proton MR‐detected metabolites have been mapped throughout the brain in a group of normal subjects using a volumetric MR spectroscopic imaging (MRSI) acquisition with an interleaved water reference. Data were processed with intensity and spatial normalization to enable voxel‐based analysis methods to be applied across a group of subjects. Results demonstrate significant regional, tissue, and gender‐dependent variations of brain metabolite concentrations, and variations of these distributions with normal aging. The greatest alteration of metabolites with age was observed for white‐matter choline and creatine. An example of the utility of the normative metabolic reference information is then demonstrated for analysis of data acquired from a subject who suffered a traumatic brain injury. This study demonstrates the ability to obtain proton spectra from a wide region of the brain and to apply fully automated processing methods. The resultant data provide a normative reference for subsequent utilization for studies of brain injury and disease. Magn Reson Med, 2009. © 2008 Wiley‐Liss, Inc.     

Proton echo-planar spectroscopic imaging of J-coupled resonances in human brain at 3 and 4 Tesla.

In this multicenter study, 2D spatial mapping of J-coupled resonances at 3T and 4T was performed using short-TE (15 ms) proton echo-planar spectroscopic imaging (PEPSI). Water-suppressed (WS) data were acquired in 8.5 min with 1-cm(3) spatial resolution from a supraventricular axial slice. Optimized outer volume suppression (OVS) enabled mapping in close proximity to peripheral scalp regions. Constrained spectral fitting in reference to a non-WS (NWS) scan was performed with LCModel using correction for relaxation attenuation and partial-volume effects. The concentrations of total choline (tCho), creatine + phosphocreatine (Cr+PCr), glutamate (Glu), glutamate + glutamine (Glu+Gln), myo-inositol (Ins), NAA, NAA+NAAG, and two macromolecular resonances at 0.9 and 2.0 ppm were mapped with mean Cramer-Rao lower bounds (CRLBs) between 6% and 18% and approximately 150-cm(3) sensitive volumes. Aspartate, GABA, glutamine (Gln), glutathione (GSH), phosphoethanolamine (PE), and macromolecules (MMs) at 1.2 ppm were also mapped, although with larger mean CRLBs between 30% and 44%. The CRLBs at 4T were 19% lower on average as compared to 3T, consistent with a higher signal-to-noise ratio (SNR) and increased spectral resolution. Metabolite concentrations were in the ranges reported in previous studies. Glu concentration was significantly higher in gray matter (GM) compared to white matter (WM), as anticipated. The short acquisition time makes this methodology suitable for clinical studies.     

Scan time reduction in proton magnetic resonance spectroscopic imaging of the human brain.

A simple technique is described for scan time reductions in proton magnetic resonance spectroscopic imaging (MRSI) of the human brain. Scan time is reduced by approximately 35% while preserving spatial resolution by reducing the field of view (FOV) and number of phase-encoding steps in the transverse direction of the brain. A multislice MRSI of the brain is demonstrated which takes approximately 20 min with a square FOV, and 13 min with a reduced FOV. The signal-to-noise ratio (SNR) in the reduced FOV scan was measured to be 15% lower than that of the full FOV scan, which is close to the expected theoretical value of 19% based on the square root of the scan time. The method can be applied with any sequence, and requires minimal software and no hardware modifications. Scan time in MRSI is minimized in this method by using FOVs no larger than the dimensions of the object to be imaged. The method may also be combined with other fast MRSI techniques to provide further scan time reductions.     

Quantitative proton MR spectroscopic imaging of normal human cerebellum and brain stem.

Quantitative, multislice proton MR spectroscopic imaging (MRSI) was used to investigate regional metabolite levels and ratios in the normal adult human posterior fossa. Six normal volunteers (36 +/- 3 years, five male, one female) were scanned on a 1.5 T scanner using multislice MRSI at long echo time (TE 280 msec). The entire cerebellum was covered using three oblique-axial slice locations, which also included the pons, mid-brain, insular cortex, and parieto-occipital lobe. Concentrations of N-acetylaspartate (NAA), choline (Cho), and creatine (Cr) were estimated using the phantom replacement technique. Regional variations of the concentrations were assessed using ANOVA (P < 0.05). High-resolution MRSI data was obtained in all subjects and brain regions examined. Metabolite concentrations (mM) (mean +/- SD) were as follows: cerebellar vermis: 2.3 +/- 0.4, 8.8 +/- 1.7 and 7.6 +/- 1.0 for Cho, Cr, and NAA respectively; cerebellar hemisphere: 2.2 +/- 0.6, 8.9 +/- 2.1, 7.5 +/- 0.8; pons 2.2 +/- 0.5, 4.3 +/- 1.1, 8.3 +/- 0.9; insular cortex, 1.8 +/- 0.5, 7.8 +/- 2, 8.0 +/- 1.1, parieto-occipital gray matter, 1.3 +/- 0.3, 5.7 +/- 1.1, 7.2 +/- 0.9, and occipital white matter, 1.4 +/- 0.3, 5.3 +/- 1.3, 7.5 +/- 0.8. Consistent with previous reports, significantly higher levels of Cr were found in the cerebellum compared to parieto-occipital gray and occipital white matter, and pons (P < 0.0001). NAA was essentially uniformly distributed within the regions chosen for analysis, with the highest level in the pons (P < 0.04). Cho was significantly higher in the cerebellum and pons than parieto-occipital gray and occipital white matter (P < 0.002) and was also higher in the pons than in the insular cortex (P < 0.05). Quantitative multislice MRSI of the posterior fossa is feasible and significant regional differences in metabolite concentrations were found.     

Short echo time proton magnetic resonance spectroscopic imaging of macromolecule and metabolite signal intensities in the human brain

A novel approach is presented for imaging macromolecule and metabolite signals in brain by proton magnetic resonance spectroscopic imaging. The method differentiates between metabolites and macromolecules by T₁ weighting using an inversion pulse followed by a variable inversion recovery time before localization and spectroscopic imaging. In healthy subjects, the major macromolecule resonances at 2.05 and 0.9 ppm were mapped at a nominal spatial resolution of 1 × 1 × 1.5 cm³ and were demonstrated to be highly reproducible between subjects. In subacute stroke patients, a highly elevated macromolecule resonance at 1.3 ppm was mapped to infarcted brain regions, suggesting potential applications for studying pathological conditions.     

Quantitative mapping of total choline in healthy human breast using proton echo planar spectroscopic imaging (PEPSI) at 3 Tesla

Purpose:

To quantitatively measure tCho levels in healthy breasts using Proton‐Echo‐Planar‐Spectroscopic‐Imaging (PEPSI).

Materials and Methods:

The two‐dimensional mapping of tCho at 3 Tesla across an entire breast slice using PEPSI and a hybrid spectral quantification method based on LCModel fitting and integration of tCho using the fitted spectrum were developed. This method was validated in 19 healthy females and compared with single voxel spectroscopy (SVS) and with PRESS prelocalized conventional Magnetic Resonance Spectroscopic Imaging (MRSI) using identical voxel size (8 cc) and similar scan times (～7 min).

Results:

A tCho peak with a signal to noise ratio larger than 2 was detected in 10 subjects using both PEPSI and SVS. The average tCho concentration in these subjects was 0.45 ± 0.2 mmol/kg using PEPSI and 0.48 ± 0.3 mmol/kg using SVS. Comparable results were obtained in two subjects using conventional MRSI. High lipid content in the spectra of nine tCho negative subjects was associated with spectral line broadening of more than 26 Hz, which made tCho detection impossible. Conventional MRSI with PRESS prelocalization in glandular tissue in two of these subjects yielded tCho concentrations comparable to PEPSI.

Conclusion:

The detection sensitivity of PEPSI is comparable to SVS and conventional PRESS‐MRSI. PEPSI can be potentially used in the evaluation of tCho in breast cancer. A tCho threshold concentration value of ～0.7 mmol/kg might be used to differentiate between cancerous and healthy (or benign) breast tissues based on this work and previous studies. J. Magn. Reson. Imaging 2012;36:1113–1123. © 2012 Wiley Periodicals, Inc.     

Achieving sufficient spectral bandwidth for volumetric 1H echo-planar spectroscopic imaging at 4 Tesla.

Complete coverage of the in vivo proton metabolite spectrum, including downfield resonances, requires a spectral bandwidth of approximately 9 ppm. Spectral bandwidth of in vivo echo-planar spectroscopic imaging (EPSI) is primarily limited by gradient strength of the oscillating readout gradient, gradient slew rate, and limits on peripheral nerve stimulation for human subjects. Furthermore, conventional EPSI reconstruction, which utilizes even and odd readout echoes separately, makes use of only half the spectral bandwidth. In order to regain full spectral bandwidth in EPSI, it has previously been suggested to apply an interlaced Fourier transform (iFT), which uses even and odd echoes simultaneously. However, this method has not been thoroughly analyzed regarding its usefulness for in vivo 3D EPSI. In this Note, limitations of the iFT method are discussed and an alternative, cyclic spectral unwrapping, is proposed, which is based on prior knowledge of typical in vivo spectral patterns.     

Fast three‐dimensional proton spectroscopic imaging of the human brain at 3 T by combining spectroscopic missing pulse steady‐state free precession and echo planar spectroscopic imaging

The combination of the principles of two fast spectroscopic imaging (SI) methods, spectroscopic missing pulse steady‐state free precession and echo planar SI (EPSI) is described as an approach toward fast 3D SI. This method, termed missing pulse steady‐state free precession echo planar SI, exhibits a considerably reduced minimum total measurement time T_min, allowing a higher temporal resolution, a larger spatial matrix size, and the use of k‐space weighted averaging and phase cycling, while maintaining all advantages of the original spectroscopic missing pulse steady‐state free precession sequence. The minor signal‐to‐noise ratio loss caused by using oscillating read gradients can be compensated by applying k‐space weighted averaging. The missing pulse steady‐state free precession echo planar SI sequence was implemented on a 3 T head scanner, tested on phantoms and applied to healthy volunteers. Magn Reson Med, 2011. © 2011 Wiley Periodicals, Inc.     

Long and short echo time proton magnetic resonance spectroscopic imaging of the healthy aging brain

PURPOSE: To investigate the relationship between subject age and white matter brain metabolite concentrations and R(2) relaxation rates in a cross-sectional study of human brain. MATERIALS AND METHODS: Long- and short-echo proton spectroscopic imaging were used to investigate concentrations and R2 relaxation rates of N-acetyl aspartate (NAA) + N-acetyl aspartyl glutamate (NAAG), choline (Cho), creatine (Cr), and myoinositol (mI) in the white matter of the centrum semiovale of 106 healthy volunteers aged 50-90 years; usable data were obtained from 79 subjects. A major aim was to identify which parameters were most sensitive to changes with age. Spectra were analyzed using the LCModel method. RESULTS: The apparent R2 of NAA and the LCModel concentration of Cr at short echo time were significantly correlated with age after multiplicity correction. Large lipid resonances were observed in the brain midline of some subjects, the incidence increasing significantly with age. We believe this to result from lipid deposits in the falx cerebri. CONCLUSION: Since only short-echo spectroscopy showed a robust relationship between Cr and subject age, and detects more metabolites than long echo time, we conclude that short-echo is superior to long-echo for future aging studies. Future studies could usefully determine whether the Cr-age relationship is due to changes in concentration, T1, or both.