SEARCH

SEARCH BY CITATION

Keywords:

  • cardiac metabolism;
  • hyperpolarized 13C;
  • metabolic imaging;
  • pyruvate;
  • bicarbonate;
  • pulse sequences;
  • RF pulses

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

Hyperpolarization of spins via dynamic nuclear polarization (DNP) has been explored as a method to non-invasively study real-time metabolic rocesses occurring in vivo using 13C-labeled substrates. Recently, hyperpolarized 13C pyruvate has been used to characterize in vivo cardiac metabolism in the rat and pig. Conventional 3D spectroscopic imaging methods require in excess of 100 excitations, making it challenging to acquire a full cardiac-gated, breath-held, whole-heart volume. In this article, the development of a rapid multislice cardiac-gated spiral 13C imaging pulse sequence consisting of a large flip-angle spectral-spatial excitation RF pulse combined with a single-shot spiral k-space trajectory for rapid imaging of cardiac metabolism is described. This sequence permits whole-heart coverage (6 slices, 8.8-mm in-plane resolution) in any plane, allowing imaging of the metabolites of interest, [1- 13C] pyruvate, [1- 13C] lactate, and 13C bicarbonate, within a single breathhold. Pyruvate and bicarbonate cardiac volumes were acquired, while lactate images were not acquired due to low lactate levels in the animal model studied. The sequence was demonstrated with phantom experiments and in vivo testing in a pig model. Magn Reson Med, 2010. © 2010 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

Hyperpolarization of spins via dynamic nuclear polarization (DNP) has been explored as a method to noninvasively study real-time metabolic processes occurring in vivo using 13C-labeled substrates (1–3). Recently, hyperpolarized 13C pyruvate has been used to characterize cardiac metabolism noninvasively in the rat (4, 5) and the pig (6). In these studies, [1- 13C] pyruvate was injected intravenously, and cardiac spectra were acquired with spatial localization to the heart either by coil location or by slice selection and single-step phase encoding of the field of view.

Spectroscopic imaging methods are typically used to resolve the spatial distribution of multiple metabolites. Conventionally, a single FID is acquired per spatial k-space point, and the full k-t dataset is then reconstructed into a chemical shift image. In a previous study by Golman et al., a noncardiac-gated 13.4 s spectroscopic acquisition with 149 excitations was used to reconstruct a single 2-cm thick slice with 7.5 × 7.5 mm2 in-plane resolution.

In cardiac applications, imaging a whole-heart volume with high spatial resolution while suppressing flow and motion, within a 20-sec cardiac-gated breath-hold, is challenging. In particular, the number of excitations is limited by the number of heart cycles in a cardiac-gated imaging sequence, and it is unfeasible to achieve this using a 3D phase-encoded spectroscopic acquisition, even with the introduction of EPI or other k-space trajectories.

As an alternative to a spectroscopic acquisition of the multiple metabolites that arise from hyperpolarized 13C pyruvate, direct imaging of each metabolite can be performed. Previously, a spectral-spatial excitation pulse has been used to selectively image the lactate resonance in the [1- 13C] pyruvate spectrum, with the resulting signal spatially encoded using a 3D multi-shot echo-planar flyback readout trajectory (7). The spectral-spatial profile of the RF pulse is designed to avoid excitation of the remaining resonances in the spectrum. In cardiac spectra containing multiple resonances of interest, such as bicarbonate, lactate, and pyruvate, alternating between the chemical shift frequencies corresponding to each metabolite allows for imaging of each individual metabolite. The relatively long Tmath image of 13C labeled compounds enables the use of a flow and motion-insensitive single-shot spiral k-space trajectory to spatially encode the signal.

In this article, we describe the development of a rapid multislice cardiac-gated spiral 13C imaging pulse sequence consisting of a large flip-angle spectral-spatial excitation RF pulse with a single-shot spiral k-space trajectory for rapid imaging of cardiac metabolism. This sequence permits whole-heart coverage (6 slices, 8.8 mm in-plane resolution) in any plane, allowing imaging of the metabolites of interest, [1- 13C] pyruvate, [1- 13C] lactate, and 13C bicarbonate, within a single breathhold. Pyruvate and bicarbonate cardiac volumes were acquired, while lactate images were not acquired due to low lactate levels in the animal model studied.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

Spiral Multislice 13C Imaging Pulse Sequence

Sequence Design

The spiral pulse sequence shown in Fig. 1a was used to acquire cardiac-gated 13C images in the short axis view (6 slices, single-shot 16,384 × 1, Tread = 64 ms, BW = 125 kHz (4 μs per point), nominal FA = 90°, slice thickness 10 mm, spacing 1 mm). Images were acquired with an FOV of 24 cm with an acquired in-plane resolution of 3.7 × 3.7 mm2. The in-plane resolution was determined by calculating the corresponding point spread function and taking the FWHM as a measure of resolution. The subsequent filtering process, described in the reconstruction section, produced an effective reconstructed resolution of 8.8 × 8.8 mm2.

thumbnail image

Figure 1. 13C spiral pulse sequence used to image multiple metabolites in the heart. a: The sequence consists of a spectral-spatial RF excitation pulse followed by a single-shot spiral readout trajectory. The k-space trajectory is also shown. b: The sequence is run in a multislice, spectrally interleaved mode to resolve volumes corresponding to bicarbonate, lactate, and pyruvate in the pig myocardium. The filled triangles indicate the cardiac triggers, and two slices are resolved 350 ms after the trigger in a 150 ms diastolic window per R-R interval.

Download figure to PowerPoint

Figure 1b shows the slice ordering used in the in vivo experiments. Slices were sequentially ordered from the atria to the ventricles. Volumes corresponding to each resonance (bicarbonate, lactate, and pyruvate) were acquired sequentially. This order was chosen based on the time-to-peak of bicarbonate and lactate in dynamic data acquired in previous experiments. The pyruvate image was acquired last so that signal from the converted metabolites was not affected. The start of the scan was chosen to coincide with the maximum bicarbonate levels as determined by the dynamic data in Fig. 5c.

Two slices were acquired per heart beat in a 160 ms window during diastole, with a trigger delay of 350 ms from the peripheral trigger placed on the pig's tail. The cardiac motion between the two acquired slices was estimated to be 0.5 cm from the cine images acquired, as detailed below. Six slices were acquired per volume. Thus, each full volume corresponding to a single resonance was acquired in three heart beats, and volumes for the three metabolites of interest acquired over 9 heart beats. This corresponded to ∼6 s, depending on the heart rate of the pig. Based on this scan time, the scan was repeated three times to yield three time points over 27 heart beats, which was placed inside a single 20 sec breathhold. To improve SNR, the three time points for each metabolite were averaged.

Linear shimming was performed over a 12-cm region encompassing the heart during the proton acquisition detailed below, and the shim values were carried over to the 13C imaging acquisition. This produced linewidths of ∼1 ppm for [1- 13C] pyruvate.

RF Pulse Design and Placement of the Spectral Passband

A spectral-spatial excitation pulse was designed according to (7). The RF pulse used was 13.6-ms long and has an excitation profile with two passbands with FWHM of 200 Hz, separated by a 1-kHz wide stopband with a suppression level of 10−3, and secondary bipolar sidelobes with suppression level of 10−2 located 500 Hz from the primary passbands. The B1 required to produce a flip angle of 90° was 0.13 G. The minimum slice thickness of the pulse, based on a maximum gradient strength of 5 G/cm and slew rate of 200 mT/m/s, was 5 mm. The secondary spatial sidelobes are located 50 cm to the side of the primary slice. The RF pulse is relatively insensitive to bulk shifts in B0 due to the 6-ppm FWHM passband used. The spectral-spatial profile of the RF pulse is linear with respect to the B1 corresponding to flip angles between 0° and 90° generated by the surface coil.

To selectively excite the resonance of interest while suppressing signal from the other resonances present, the location of the passbands of the excitation pulse must be carefully chosen (Fig. 2). Specifically, the signal from injected hyperpolarized 13C pyruvate is typically an order of magnitude higher than its downstream products, and needs to be suppressed to avoid image artifacts.

thumbnail image

Figure 2. Placement of the spectral passband of the spectral-spatial RF excitation pulse. The dualband excitation pulse used has an excitation profile with two passbands, a stopband width of 1 kHz with a suppression level of 10−3, and secondary sidelobes located 500 Hz from the primary passbands. This profile influences which passband is used to excite each resonance in order to avoid off-resonance excitation. The bicarbonate resonance is excited using the negative passband, while the lactate and pyruvate resonances are excited using the positive passband.

Download figure to PowerPoint

For the spectrum that arises from [1- 13C] pyruvate, the chemical shifts relative to pyruvate are 385, 270, 185, 0, and −310 Hz, for lactate, pyruvate hydrate, alanine, pyruvate, and bicarbonate, respectively. Based on the excitation profile, the passbands used to excite lactate and bicarbonate are chosen so that pyruvate lies in the stopband region with a suppression level of 10−3. Pyruvate is excited using the positive (left) passband, which is necessary to place bicarbonate in the suppression region (see Fig. 2).

Spiral Reconstruction

The reconstruction process involving an automatic off-resonance correction algorithm is outlined in Fig. 3. The k-space data were smoothed using an exponential filter with time constant 8 ms in the time domain, gridded onto a 80 × 80 Cartesian grid using a Kaiser-Bessel kernel with a kernel width of 1.5 and an overgrid factor of 2, and then 2D Fourier transformed to obtain the image (8). The effective reconstructed resolution was determined to be 8.8 × 8.8 mm2 using the FWHM of the point spread function as described previously.

thumbnail image

Figure 3. Flowchart illustrating the reconstruction process for the spiral 13C data. The raw data are first reconstructed at several frequencies given by fmin,fminf,…,fmax using a multifrequency interpolation (MFI) scheme. Next, the auto-focusing objective function given by Eq. 1 is computed, and an estimated coarse field map is computed by minimizing the objective over the range of frequencies. This information is used to piece together the final image on a pixel-by-pixel basis.

Download figure to PowerPoint

Off-resonance blurring due to B0 inhomogeneity was corrected using an automatic field-map estimation algorithm (9, 10). Briefly, the acquired data were reconstructed at several frequencies covering the estimated range of off-resonance using multifrequency interpolation (11). Next, incidental phase across the image was removed by computing a low-pass filtered version of the original image at each frequency. After this phase removal, an in-focus image with no off-resonance blurring will be purely real. A low resolution field map at each point can be estimated by determining the frequency, which minimizes an objective function involving the imaginary part of the image, as given in Eq. 1.

  • equation image(1)

Here, I(x, y, Δω) is the image reconstructed at frequency Δω, α is chosen to be 1, and A(x,y) is a local integration window at location (x,y). The size of this window is inversely related to the frequency range over which the minimization is performed, and determines the resultant resolution of the estimated field map. In the final reconstruction, this window is chosen to be 10 × 10 cm2. Once the field map has been computed, the final image is pieced together pixel-by-pixel by retrieving the corresponding pixel value at the estimated field map value.

Phantom Imaging

To test the image quality achievable with the new sequence, images were acquired on a phantom with spheres containing thermally polarized 13C-labeled lactate, alanine, and glycine (8 slices, single-shot 16,384 × 1, Tread = 64 ms, BW = 125 kHz (4 μs per point), nominal FA = 90°, FOV 48 cm, in-plane reconstructed resolution 1.6 × 1.6 cm2, slice thickness 10 mm, spacing 1 mm, TR 700 ms, NEX 128, scan time 4.5 min). The positions of the spheres vary as a function of the slice location in the phantom. The resolution was chosen to improve the SNR when imaging the thermally polarized 13C-labeled phantom. A linear transmit/receive dual-tuned 1H/13C birdcage coil (Magvale, Sunnyvale, CA) was used for imaging. The chemical shifts relative to 13C lactate are 0, −225, and −320 Hz for lactate, alanine, and glycine, respectively. A similar passband placement to the in vivo situation was followed to avoid unwanted excitation of each resonance. The change in the lactate to alanine chemical shift relative to the in vivo shift is specific to this phantom. Corresponding 1H GRE images were acquired in the same location (FOV 16 cm, 256 × 128 matrix, slice thickness 2 mm, spacing 9 mm).

Animal Model

All animal experiments were carried out under a protocol approved by the institutional animal care and use committee. Specific pathogen free (SPF) pigs (25 kg) were fasted for one day prior to scanning due to concerns over asphyxiation while under in the scanner. To raise blood glucose levels during the scan to increase PDH activity, 1 L electrolyte solution (Life Brand, Toronto) containing 25 g glucose was given orally 1 hour prior to anesthesia.

Ketamine anesthesia was used for induction, and isoflurane at 2.5% in oxygen was used to maintain anesthesia for the duration of the experiment. Blood oxygen saturation and heart rate was monitored using a peripheral pulse oximeter placed on the tail. The temperature of the pigs under anesthesia was maintained using a blanket during the duration of the experiment. The pigs were placed on a ventilator to maintain respiration at a constant rate of 24 breaths per minute. Respiration was monitored using a respiratory belt placed around the torso of the pig. Breathholding to eliminate respiratory motion was performed by disabling the ventilator.

Experimental Setup

All experiments were performed on a GE MR750 3T MR scanner. For 13C imaging, a custom-built transmit/receive 13 cm 13C surface coil with 1H blocking was placed over the pig chest. Phantom experiments using an oil phantom were used to calibrate the required transmit power to achieve the desired flip angle in the heart, which is located ∼ 4 cm from the coil.

For anatomical landmarking, cardiac-gated breath-held SSFP CINE images were acquired in the short-axis view (TR = 4.2 ms, TE = 1.8 ms, FOV 24 cm, slice thickness 5 mm, spacing 5 mm, matrix size 224 × 224) using a separate 1H surface coil. The pig was not moved when the coil was replaced between proton and carbon acquisitions.

Polarization and Injection Protocol

Dynamic nuclear polarization (DNP) and dissolution (1) was used to produce hyperpolarized [1- 13C] pyruvate in solution. For each experiment, 105 μL of [1- 13C] pyruvic acid (99%, Isotec, Miamisburg, OH) with 15 mM OX63 trityl radical (Oxford Instruments, Abingdon, UK) and 1 mM of ProHance (Gadoteridol; Bracco Diagnostics, Inc., Princeton, NJ) was hyperpolarized using a HyperSense DNP polarizer (Oxford Instruments, Abingdon, UK). Six milliliters of NaOH/Tris/EDTA solution was used to dissolve the sample to obtain a 250 mM pyruvate solution. This solution was then diluted with normal saline for a final pyruvate concentration of 83 mM. 15 mL of the prepolarized [1- 13C] pyruvate in solution was injected into the animal in each 13C experiment. The [1- 13C] pyruvate polarization in solution was ∼15–20%, and the pH of the solution was 7.55 ± 0.07 in these studies. The transfer time of the hyperpolarized solution between the polarizer and the scanner was ∼15 sec.

Fifteen milliliters of hyperpolarized 83 mM [1- 13C] pyruvate was injected over 15 sec into the right ear vein. Three milliliters of saline solution was then injected over 3 sec to flush the dead volume in the injection line. The pyruvate dose used in these experiments was 1.25 mmol, corresponding to a dose of 0.05 mmol/kg for a 25 kg pig. For the 13C dynamic studies, the scan was started at the beginning of the injection. For the 13C imaging studies involving either flyback echo-planar MRSI or the spiral readout, a breathhold was started at 22 sec after the start of the injection by disabling the ventilator, and the scan was started 25 sec after the start of the injection.

A total of three animals were scanned in this study. Two dynamic acquisitions were performed, on one fasted pig and one oral glucose-loaded pig fed the electrolyte solution prior to the scan. On the oral glucose-loaded pig, an echo-planar MRSI dataset was acquired during the same session. One multislice spiral dataset was acquired on a separate, oral glucose-loaded pig.

Dynamic 13C Acquisition

To determine the appropriate timing for subsequent experiments, a cardiac-gated pulse-and-acquire pulse sequence (slice thickness 10 cm, nominal FA = 10°) was used to resolve dynamic metabolism. Spatial localization was provided by a combination of the small tip-angle slice-selective sinc excitation pulse and the placement of the 13C surface coil over the surface of the heart. One spectrum (2048 spectral points, 5 kHz bandwidth) was acquired cardiac gated to every 3 or 4 R-R intervals, depending on heart rate, to acquire one time point approximately every 2 s.

Flyback EPI CSI

For comparison with the new pulse sequence, and to determine the spectral quality that can be achieved at different locations within the heart, a cardiac-gated 13C chemical shift imaging (CSI) acquisition was performed. This consisted of a small flip-angle sinc excitation (nominal FA 20°) followed by an echo-planar flyback readout (16 pts per lobe, 128 lobes, spectral BW 446 Hz, Tread 280 ms) (12). An interleaved strategy was applied in alternate TRs by shifting the flyback readout by half the time between each echo-planar lobe to double the spectral bandwidth to 892 Hz. This bandwidth was sufficient to cover the chemical shift between lactate and bicarbonate. The acquired matrix size was 16 × 16 spatial points with 256 spectral points, with an in-plane resolution of 6 mm and a 1 cm slice thickness, giving a 96 × 96 mm2 field of view. The cardiac-gated scan required 32 heart beats, and the total scan time was between 18 and 24 sec, given a physiological heart rate range of 80 to 110 beats per minute.

MRSI Analysis

Dynamic spectra were apodized in the time domain using a 10 Hz gaussian filter prior to a one-dimensional Fourier transform. The resulting spectra were subsequently corrected using frequency-dependent linear phase and baseline correction to account for phase accrual during the RF excitation. Peak areas in the real channel were quantified using SAGE (GE Healthcare). For each dynamic data set, relative bicarbonate production was measured by calculating the maximum bicarbonate peak area to maximum pyruvate peak area ratio in the real channel.

The 13C flyback echo-planar MRSI data were analyzed using a combination of Matlab and SAGE. The data were deinterleaved in the time domain (to produce the full spectral bandwidth), and apodized using a 10 Hz gaussian filter. The spatial k-space data were apodized using a Fermi filter and zero-filled from 16 × 16 to 32 × 32 points.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

Figure 4 shows 13C lactate, alanine, and glycine phantom images acquired using the multislice sequence overlaid on corresponding 1H slices. Each image is cropped to 8 × 8 cm2 with an in-plane resolution of 1.6 × 1.6 cm2. The slight misregistration is presumed to be due to the coarse resolution of the 13C images used.

thumbnail image

Figure 4. Images acquired using the multi-slice cardiac-gated spiral sequence showing spatial distribution of 13C lactate, alanine, and glycine in a small phantom. Each column depicts an adjacent slice in the phantom. Each image is cropped to 8 × 8 cm2. The resolution of the overlaid reconstructed images is 1.6 × 1.6 cm2 in-plane with a 1-cm slice thickness. The color scale ranges from 30 to 100% of the image intensity maximum.

Download figure to PowerPoint

Figure 5a and b demonstrate the apparent modulation of PDH activity between the glucose-loaded and fasted state of the animal. In the overnight-fasted pig, the measured maximum bicarbonate to maximum pyruvate ratio (BPR) was 2 orders of magnitude lower than in the pig given an electrolyte-sugar solution to drink one hour prior to the scan. Figure 5c shows a representative cardiac-gated dynamic 13C MRS dataset localized to the heart by the combination of a 10-cm slice selective RF excitation pulse as well as by the sensitive region of the 13C surface coil used. Spectra were acquired every 3 to 4 R-R intervals, depending on heart rate, for an approximate TR of 2 sec.

thumbnail image

Figure 5. Cardiac-gated data showing dynamics of metabolites localized to the pig myocardium. a: Representative spectrum from the maximum bicarbonate frame in a fasted pig. The maximum bicarbonate peak area to maximum pyruvate peak area ratio (BPR) was calculated to be 2.4 × 10−4. b: Representative spectrum from the maximum bicarbonate frame in an oral glucose loaded pig. The BPR was calculated to be 0.07. c: Time course of peak areas of pyruvate, bicarbonate, lactate, and alanine resonances acquired every 4 R-R intervals in the oral glucose loaded pig.

Download figure to PowerPoint

Figure 6 contains cardiac-gated CSI data showing the spatial distribution of pyruvate and bicarbonate with 6 × 6 mm2 in-plane resolution in a single 10-mm thick axial slice through the pig myocardium. The pyruvate image shows pyruvate localized mainly to the left ventricle, but also to the right ventricle and the internal thoracic arteries. The bicarbonate image shows bicarbonate produced in the myocardium. The SNR of the images drops off near the posterior myocardium, which is related to reduction in B1 due to the surface transmit/receive coil used for imaging. The inset spectrum is a sum over all magnitude spectra from all voxels, demonstrating the uniformity of the shim over the heart. The linewidth of both pyruvate and bicarbonate is ∼1 ppm FWHM. The slightly broader bicarbonate resonance is presumably due to susceptibility differences from the myocardium being closer to the air-tissue interface with the lung. No lactate resonance appears in the sum spectrum at its expected chemical shift.

thumbnail image

Figure 6. Cardiac-gated CSI data showing spatial distribution of (a) pyruvate and (b) bicarbonate in a single 10-mm thick axial slice through the pig myocardium. The inset spectrum is a sum over all magnitude spectra from all voxels. The scale bar indicates 2 cm. The color scale for both images ranges from 15 to 100% of the image intensity maximum.

Download figure to PowerPoint

Figure 7 shows in vivo data acquired using the multislice cardiac-gated spiral sequence in a short-axis view of the heart reconstructed using the automatic off-resonance correction algorithm described in the “Methods” section. Each image is cropped to 12 × 12 cm2 with an in-plane resolution of 8.8 × 8.8 mm2. The distribution of pyruvate is again within the chambers of the heart, while bicarbonate is localized to the myocardium. The lactate image generated by this technique was below the detection limit, which is similar to the CSI data in Fig. 6. As seen in Fig. 6, the SNR of the images is reduced in the posterior myocardium. However, the geometry of the short axis view brings some of the slices closer to the coil than others, and this is seen in the increased SNR in the central slices of the bicarbonate images.

thumbnail image

Figure 7. In vivo data acquired using the multislice cardiac-gated spiral sequence showing spatial distribution of metabolites in a short-axis view of the heart. Bicarbonate, lactate, and pyruvate volumes were acquired over 9 heart beats, and the sequence was repeated for three time points. Each image is cropped to 12 × 12 cm2. The resolution of the overlaid reconstructed images is 8.8 × 8.8 mm2 in-plane with a 1-cm slice thickness. The entire scan was completed within 18 sec. The colour scale ranges from 15 to 100% of the image intensity maximum.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

The multislice sequence used here provides a simple method to extend 13C imaging from a single slice to full volumetric coverage of the heart. Multislice imaging allows for slice-by-slice shimming to improve image quality and B0 homogeneity (13).

In cardiac imaging, the number of excitations must be low in order to allow for cardiac gating. In addition, the sequence used must be insensitive to flow and motion present in cardiac imaging. In a previous article by Cunningham et al. describing the design of a spectral-spatial RF excitation pulse for lactate imaging, a 3D flyback multi-shot EPI readout trajectory was used to spatially encode a 3D volume over 64 excitations. This strategy is infeasible for cardiac imaging because of the limited number of excitations available as well as the large gradient moments of the flyback EPI trajectory. This motivates the use of a flow-insensitive single-shot spiral readout trajectory, enabling imaging of a single slice with a single excitation. Reconstruction of the non-Cartesian spiral data via gridding is well understood and commonly used (8).

By enabling cardiac gating, the sequence described here produces images with improved spatial resolution over non-gated techniques, such as that described in Golman et al. (6). Nominally, the in-plane resolution reported in that study was 7.5 × 7.5 mm2. The actual resolution of the Golman study is degraded by cardiac motion. The 8.8 × 8.8 mm2 in-plane resolution used in the sequence in this paper allows for resolution of bicarbonate in various myocardial structures, including both the left ventricle wall as well as part of the right ventricle wall (Fig. 7).

Each volume was acquired with slices ordered sequentially from the atria to the ventricles and an entire volume was acquired for each metabolite, over three heart beats, before moving on to the next. The acquisition order used provides the most temporal registration between the slices in each volume. This could be extended into a dynamic multislice acquisition by repeating this acquisition several times, with the metabolic images interleaved to yield a time course of metabolites. This ordering also has the advantage of removing misregistration issues by placing each acquired slice at the same cardiac phase in the corresponding metabolic images. Alternately, the acquisition could be ordered with each slice being acquired three times before moving to the adjacent slice. This would temporally register each slice, making it possible to directly compare metabolite concentrations at each time point.

One concern with the multislice nature of the sequence is the impact of flow or perfusion on the images acquired. We note that movement of spins due to blood flow in the chambers may lead to variation in signal from slice to slice, and this may impact the images acquired of the pyruvate bolus. Presumably, intracellular bicarbonate and lactate generated from the injected pyruvate in adjacent slices do not move appreciably due to perfusion over the course of the scan.

The flip angle used in this sequence is nominally set to 90° in the heart, ∼4 cm from the 13C transmit/receive surface coil used. The large flip angle used for this acquisition was chosen to maximize the SNR of the resulting images. The temporal ordering of the metabolic images (bicarbonate, lactate, pyruvate) was chosen to avoid saturation of the hyperpolarized signal by the large flip angle in the downstream metabolites. The flip angle could be reduced to allow for the possibility of dynamic imaging. The feasibility of this approach would depend on the available hyperpolarized signal. Alternately, because the hyperpolarized magnetization is being transferred from pyruvate, a large flip angle excitation could be used to selectively image bicarbonate and lactate dynamically, with the 13C magnetization being replenished by metabolic conversion. The metabolic time course will vary depending on the flip angle used, and a flip angle correction scheme is required if kinetic modeling is used to extract metabolic flux information from the data.

However, in the hyperpolarized situation, accurate B1 calibration is challenging due to T1 relaxation and the limited magnetization available. Also, the nonuniform B1 profile of the surface coil guarantees that the flip angle will deviate from the desired 90° excitation in the heart. If the flip angle is indeed not exactly 90°, the remaining longitudinal magnetization can be used by repeating the sequence several times. The acquired data can then be summed prior to reconstruction to improve image quality. Although this potentially changes the contrast of the images, the increased signal is beneficial when using the automatic off-resonance correction algorithm to estimate the field map. The signal dropoff in the posterior region of the myocardium is presumed to be due to increase in distance from the surface transmit-receive coil. This issue can be avoided with the use of a volume transmit-receive coil with a more uniform B1.

The multislice strategy taken in this sequence is extremely flexible and can be adapted to a variety of applications. As described above, the specific order of spatial and metabolic slices may be useful in certain situations. For example, limited lactate is observed in the healthy hearts scanned in this study. When there is a single metabolite of interest, the acquisition window can be used to more efficiently image that resonance, by obtaining dynamic information and improving the SNR and spatial homogeneity of the resulting metabolite image. In this article, we have chosen the most general scheme that allows for acquisition of all relevant metabolites when the underlying metabolic profile is unknown.

The sequence described in this article can acquire one slice per resonance in 80 ms with a single spectral-spatial excitation. This strategy is tailored to cardiac imaging where the number of excitations is limited by the number of cardiac cycles and the length of the breathhold window. Alternative techniques reported in the literature include conventional 2D chemical shift imaging (6), multiband 3D EPSI chemical shift imaging (14), multislice spiral chemical shift imaging (15–17), and 3D spectral-spatial multiecho bSSFP imaging (18). In cardiac-gated imaging, these methods suffer from a prohibitive number of excitations and imaging readouts which are sensitive to cardiac flow and motion. In addition, the single-shot technique described here is unique in that the data acquired is temporally registered to a single heart cycle, which is a feature not present in multi-shot or 3D techniques.

The automatic off-resonance correction algorithm used in the reconstruction estimates a coarse field map according to the auto-focusing criterion given by Eq. 1. The resolution of this field map is dependent on the size of the local integration window used to compute the autofocusing objective function, as well as the frequency range to minimize over. This window is chosen to minimize spurious minima, which arise from the spreading out of the off-resonance PSF into a large region.

To improve on the resolution of the estimated field map, semiautomatic methods involving either a separately or simultaneously acquired B0 map can constrain the frequency range to minimize over. This would reduce the size of the integration window required in Eq. 1. In addition, a B0 map would allow for correct setting of transmit and receive frequencies on a slice-by-slice basis. Potentially, this information could be combined with slice-by-slice shimming to further improve B0 homogeneity over the heart.

Although acquiring a 13C field map in vivo is challenging, a proton field map can easily be acquired and converted to carbon by scaling by the γ13C1H ratio. An alternate possibility is to use an external 13C- and 1H-containing sample to calibrate the two maps.

The objective of the oral glucose loading protocol is to manipulate the metabolic state of the animal towards a fed state following an overnight fast. We note that this protocol is essential to increase apparent PDH-mediated bicarbonate production to levels which can be imaged using the sequence described in this article. Our observations that the dietary status of the animal influences bicarbonate production from 13C pyruvate agrees with previous data demonstrated by Schroeder et al. in small animals (4). Similarly, in human cardiac metabolic studies using 18FDG PET and SPECT, spatial heterogeneity in uptake has been observed in the fasted state (19–21). Oral glucose loading as well as hyperinsulinemic-euglycemic clamping have been suggested as methods to improve homogeneity of uptake in the myocardium. The dose of oral glucose used in this study was 25 g in 1 L of electrolyte solution, which is similar to the weight-adjusted oral glucose load recommended for use in human 18FDG studies (22, 23). Alternatively, each 13C pyruvate injection could be preceded by injection of a bolus of insulin to prime the subject for uptake and conversion of the hyperpolarized substrate by increasing the activity of PDH in the heart. The improved spatial homogeneity in uptake would presumably facilitate imaging of metabolic differences between regions of the myocardium after an ischaemic event.

Blood oxygenation (95–99%) and heart rate (80–90 bpm) were monitored over the course of the experiment and were not significantly affected by the pyruvate injections. The administered 13C pyruvate dose in this study was 0.05 mmol/kg. For a typical swine blood volume of 60 mL/kg, this dose produces a plasma pyruvate concentration of 800 μM. Although the administered concentration is high relative to the endogenous plasma pyruvate concentration of 80–300 μM (24), other studies investigating the effect of administered dose in rats have demonstrated saturation of myocardial PDH-mediated reactions at concentrations far in excess of those reported here (25). In addition, the dose used in this study is six times less than the dose reported in Golman et al. (6).

The lactate resonance is observed in the dynamic spectra in healthy controls, but is absent in both the spatially localized chemical shift images and single-metabolite multislice spiral images of the heart. Presumably, the imaging data indicates that in the healthy control hearts imaged in this study, the majority of the administered pyruvate enters the TCA cycle for oxidative phosphorylation. These results appear to disagree with previous data in Golman et al. (6), in which lactate is spatially localized to the myocardium in conventional single-slice nontriggered CSI data.

The dynamic acquisition is performed using a slice-selective sinc excitation to localize the signal to the myocardium, with the centre frequency set to the pyruvate frequency. This leads to a chemical shift misregistration artifact in which the lactate signal arises from a shifted slice containing the liver, inferior to the myocardium. The dose used in this study (0.05 mmol/kg) is also lower relative to the Golman study (0.3 mmol/kg). Furthermore, myocardial uptake and metbolism has been shown to depend on the concentration of the hyperpolarized substrate (25). The Golman study also reports using intravenous lactated Ringer's solution for hydration. The high concentration of lactate present in this solution (28 mM) would presumably increase myocardial lactate concentration. Combined with reports that a large portion of the 13C lactate signal seen in the hyperpolarized 13C pyruvate experiment is due to exchange of the hyperpolarized label (3) suggests that the metabolic state of the heart in the Golman study is significantly different from the cardiac model described in this article.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

We have developed a spiral pulse sequence for rapid multislice imaging of cardiac hyperpolarized metabolic products. In this sequence, a large flip-angle spectral-spatial RF pulse is used to excite a single metabolite of interest, and the resulting transverse magnetization is read using a single-shot spiral k-space trajectory. This sequence permits whole-heart coverage (6 slices, 8.8 mm in-plane resolution) in any plane, allowing imaging of the metabolites of interest, [1- 13C] pyruvate, [1- 13C] lactate, and 13C bicarbonate, within a single breathhold. Pyruvate and bicarbonate cardiac volumes were acquired, while lactate images were not acquired due to low lactate levels in the animal model studied. The new sequence is anticipated to be useful in the noninvasive monitoring of changes in the spatial distribution of metabolites in disease.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES

The authors are grateful for support from the Canadian Institutes of Health Research.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES