Cardiac T₂* Measurement of Hyperpolarized ¹³C Metabolites using Metabolite-Selective Multi-Echo Spiral Imaging

Abstract

Purpose:

Noninvasive imaging with hyperpolarized pyruvate can capture in vivo cardiac metabolism. For proper quantification of the metabolites and optimization of imaging parameters, understanding MR characteristics such as T₂*s of the hyperpolarized signals is critical. This study is to measure in vivo cardiac T₂*s of hyperpolarized [1-¹³C]pyruvate and the products in rodents and humans.

Methods:

A dynamic ¹³C multi-echo spiral imaging sequence that acquires [¹³C]bicarbonate, [1-¹³C]lactate, and [1-¹³C]pyruvate images in an interleaved manner was implemented for a clinical 3T system. T₂* of each metabolite was calculated from the multi-echo images by fitting the signal decay of each region of interest mono-exponentially. The performance of measuring T₂* using the sequence was validated using a ¹³C phantom, then with rodents following a bolus injection of hyperpolarized [1-¹³C]pyruvate. In humans, T₂* of each metabolite was calculated for left ventricle (LV), right ventricle (RV) and myocardium.

Results:

Cardiac T₂*s of hyperpolarized [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate in rodents were measured as 24.9 ± 5.0, 16.4 ± 4.7, and 16.9 ± 3.4 ms, respectively. In humans, T₂* of [1-¹³C]pyruvate was 108.7 ± 22.6 ms in LV and 129.4 ± 8.9 ms in RV. T₂* of [1-¹³C]lactate was 40.9 ± 8.3, 44.2 ± 5.5, and 43.7 ± 9.0 ms in LV, RV and myocardium, respectively. T₂* of [¹³C]bicarbonate in myocardium was 64.4 ± 2.5 ms. The measurements were reproducible and consistent over time after the pyruvate injection.

Conclusions:

The proposed metabolite-selective multi-echo spiral imaging sequence reliably measures in vivo cardiac T₂*s of hyperpolarized [1-¹³C]pyruvate and products.

Introduction

MRI or MRS with hyperpolarized (HP) [1-¹³C]pyruvate can assess metabolic state of the heart. In particular, detection of HP [¹³C]bicarbonate is a direct measure of pyruvate dehydrogenase (PDH) activity and thus an important indicator of cardiac metabolism and function. For instance, myocardial [¹³C]bicarbonate production is sensitive to several pathophysiological states, including myocardial infarction,¹ diabetic cardiomyopathy,² and heart failure.^3,4 Consequently, cardiac imaging with HP [1-¹³C]pyruvate is being translated to humans.^5–7 Nonetheless, quantitative analysis of HP signals in the heart need to be established.

Cardiac imaging of HP signals is particularly challenging due to the cardiac motion and the rapid circulation of blood through the heart as well as the dynamic and transient nature of HP signals. To achieve rapid k-space sampling within a specific cardiac phase (e.g., end diastole), gradient recalled-echo (GRE)-based imaging sequences with spectral-spatial (spsp) radiofrequency (RF) pulses that excite one metabolite at a time are commonly used rather than spectroscopic imaging methods. Both spiral^5,8 or echo-planar imaging (EPI)^9,10 readouts were proposed for cardiac imaging previously.

However, cardiac HP acquisition schemes are still considered suboptimal,¹¹ and the HP ¹³C images are sensitive to acquisition parameters such as echo times, B₀ field inhomogeneity^12,13 and blood flow.¹⁴ For proper assessment of HP metabolites, in vivo MR characteristics such as the T₂*s of the HP ¹³C-metabolites need to be understood. Moreover, since T₂*s play critical roles in determining acquisition parameters, the knowledge of the T₂* of HP [1-¹³C]pyruvate and its products can be used in designing the k-space sampling trajectories for optimal signal-to-noise ratio (SNR) and spatial resolution. Previous studies estimated T₂*s of in vivo HP metabolites from the spectral linewidths using MR spectroscopy.¹³ However, this approach measures aggregated T₂* information without assessing spatial and compartmental heterogeneity. A spectroscopic imaging approach, such as chemical shift imaging (CSI) or echo-planar spectroscopic imaging (EPSI), can be used to obtain spatial T₂* information of metabolites of interest by mapping the spectral linewidths,¹⁵ but is not adequate for HP ¹³C cardiac imaging due to the transient characteristics of HP signals and the relatively long acquisition time.

In this study, we propose a dynamic metabolite-selective multi-echo spiral imaging (MESI) ¹³C sequence for imaging cardiac T₂*s of HP [1-¹³C]pyruvate and the products in vivo. The feasibility and robustness of the method were validated using a ¹³C phantom and tested with rodents in vivo. The sequence was further applied to healthy volunteers, and T₂* relaxation times of HP pyruvate, lactate, and bicarbonate were measured in left ventricle (LV), right ventricle (RV) and myocardium (Myo).

Methods

MR Scanner and RF coils

All the MR studies were performed using a clinical 3T wide-bore (diameter ⊘ = 70 cm) MRI scanner (750w Discovery, GE Healthcare, Waukesha, WI, USA). A ¹H/¹³C dual-tuned quadrature transmit/receive birdcage RF coil (inner diameter ⊘_inner = 60 mm, GE Healthcare) was used for phantom and rat studies. For human studies, a two-loop ¹³C transmit-receive Helmholtz coil was used (⊘ = 20 cm; PulseTeq Limited, Chobham, Surrey, UK). One loop was positioned over the anterior chest and the other was placed below the scapula of the human subject.

Hyperpolarization

A SPINlab^™ system (GE Healthcare) that operates at ~0.8 K in a 5T magnet was used for the dynamic nuclear polarization (DNP) of [1-¹³C]pyruvate. For animal studies, a 35-μL sample of 14-M [1-¹³C]pyruvic acid mixed with OX063 trityl (15 mM) was prepared in a research fluid path for each dissolution. After 3 – 4 hrs of polarization, the sample was dissolved with hot solvent (16 mL saline with 0.1-g/L Na₂EDTA) then mixed with pH-neutralization media (750 μL with 0.72-M NaOH). The final HP solution (~6.5 mL) that contained ~70-mM of HP [1-¹³C]pyruvate was injected through the tail vein up to 4.0 mL (0.875 mmol/kg body weight) with an injection rate of 0.25 mL/s.

For human studies, [1-¹³C]pyruvic acid (Sigma Aldrich, St. Louis, MO), produced in accordance with Good Manufacturing Practice (GMP) regulations, was prepared in clinical fluid paths as described previously (0.40 mL/kg body weight of 250-mM HP [1-¹³C]pyruvate solution).¹⁶ HP [1-¹³C]pyruvate was dissolved after 3 – 4 hrs of polarization. Pyruvate concentration, pH, temperature, volume and radical concentration of the HP solution was examined by a dedicated quality control (QC) device (GE Healthcare). Immediately after passing the QC, sterility of the sample was confirmed prior to injection as previously described.¹⁷ The HP solution was transported to the magnet and administered intravenously to subjects (injection rate = 5 mL/s), followed by a 25-mL saline flush.

RF Pulse Design and Metabolite-Interleaved Multi-Echo Spiral Imaging Sequence

A spsp RF pulse was designed to have a full width at half maximum (FWHM) of 134.4 Hz using the Spectral-Spatial RF Pulse Design Package¹⁸ to selectively excite HP [¹³C]bicarbonate, [1-¹³C]lactate, or [1-¹³C]pyruvate, depending on the acquisition frequency, Fig. 2. Unlike other spsp RF pulses used in previous HP ¹³C cardiac studies,^5,8,9 this RF pulse was specifically designed and tested for the wide-bore system (GE 750w), which has limited gradient performance (maximum gradient [G_max] = 33 mT/m, maximum slew rate [Slew_max] = 120 T/m/s), and operates with minimum gradient requirements (G_max < 25 mT/m and Slew_max < 100 T/m/s). The spsp excitation profiles, simulated using MATLAB (MathWorks, Natick, MA, USA) and tested in the scanner, are shown in Fig. 2(B). The simulated transverse signal in log scale at the center of the image slice (horizontal dotted white line) showed 90° excitation in the passband and less than 1% excitation in the stopband. Metabolite-selective MESI scheme was implemented to sequentially acquire metabolite maps using the spsp RF pulse in an interleaved manner by shifting the transmit and receive frequencies.

Figure 2. Spectral-spatial RF pulse and profiles.

(A) Spectral-spatial RF pulse and slice-selective gradient waveforms used for bicarbonate, lactate, and pyruvate excitation, and (B) the simulated (top left) and tested (bottom left) excitation profile. The simulated spectral profile at the center of the slice (horizontal dotted white line in the top left figure) in log scale is shown at the top right, and the simulated slice profile at the center of the passing band (vertical dotted white line in the top left figure) is shown at the bottom right.

Phantom Validation

The sequence was tested with a gadolinium-doped (1-mM) spherical [¹³C]bicarbonate phantom (1.0 M, ⊘ = 3 cm). ¹H MRI was acquired with two-dimensional (2D) T₁-weighted fast spin echo (FSE) sequence (repetition time [TR] = 700 ms, echo time [TE] = 12 ms, flip angle [FA] = 90°, field of view [FOV] = 60 mm × 60 mm, slice thickness [ST] = 4 mm, spatial resolution = 0.23 mm × 0.23 mm). For ¹³C MRI, five-echo images of each metabolite were acquired following the spsp RF excitation with two-shot spiral readouts (FOV = 60 mm × 60 mm, spatial resolution = 3 mm × 3 mm, FA = 90°, ST = 25 mm, TR = 3 s, TE of 1^st echo = 18.01 ms, ΔTE = 11.62 ms, readout length = 58.10 ms/arm, receiver spectral bandwidth = 85.06 Hz, #averages = 128). The decay of mean ¹³C signal within the phantom was fitted to a mono-exponential function to measure the T₂* of [¹³C]bicarbonate. From the same phantom, ¹³C spectrum was acquired using a pulse-and-acquire sequence (FA = 90°, ST = 25 mm, TR = 3 s, receive spectral bandwidth = 2000 Hz, total #spectral points = 2048, #averages = 128). The T₂* from the MESI was compared to the linewidth from the spectroscopic data. In the frequency domain, the ideal MRS signal can be described by the Lorentzian line shape,¹⁵ which is in the form as below in magnitude:

$$\mathit{S}(\mathit{f})=\frac{\mathit{H}}{\sqrt{{\mathit{W}}^2+{\left({\mathit{F}}_0-\mathit{f}\right)}^2}}$$ #(1)

where H is the maximum of the signal in magnitude, F₀ is the center frequency, and W is the signal’s FWHM ($\mathit{W}=\frac{1}{\mathit{\pi }{\mathit{T}}_2^{\mathit{*}}}$). T₂* of the phantom was calculated by fitting the spectrum to Equation (1).

T₂* Measurement of Rat Heart

Cardiac T₂*s of HP ¹³C-metabolites were measured from healthy male Wistar rats (n = 3, body weight = 556.7 ± 64.7 g). First, 2D axial T₂-weighted FSE ¹H images were acquired for anatomical reference (TR = 5000 ms, TE = 63 ms, FA = 160°, FOV = 80 mm × 80 mm, ST = 2 mm, spatial resolution = 0.63 mm × 0.63 mm). Each rat was imaged twice using the ¹³C MESI sequence (single shot, FOV = 80 mm × 80 mm, spatial resolution = 8 mm × 8 mm, ST = 25 mm, TR = 213 ms, #echo = 10, TE of 1^st echo = 15.19 ms, ΔTE = 5.98 ms, readout length = 59.76 ms/arm, receive spectral bandwidth = 167.34 Hz, FA = 90° for bicarbonate, 90° for lactate, and 5° for pyruvate, #timepoint = 16) with two consecutive injections of HP [1-¹³C]pyruvate solution. Images of [¹³C]bicarbonate, [1-¹³C]lactate, and [1-¹³C]pyruvate were acquired from an axial plane containing the heart after a 15-s delay from the start of each injection. The minimum time interval between the injections was 20 min. The acquisition scheme is summarized in Fig. 1. Note that the ECG gating was used only for humans. All procedures were approved by the local Institutional Animal Care and Use Committee.

Figure 1. Metabolite-interleaved ¹³C multi-echo spiral imaging sequence.

[¹³C]Bicarbonate, [1-¹³C]lactate and [1-¹³C]pyruvate are sequentially excited using a spectral-spatial RF pulse and imaged with multiple single-shot spiral readouts.

Cardiac T₂* Measurement of Healthy Human Subjects

Three healthy subjects were recruited for the study (2 female, 1 male, age = 38 – 48 years, Table 1). After positioning the subject in the magnet, horizontal long-axis (HLA), vertical long-axis (VLA) and short-axis (SA) images were acquired using a ¹H balanced steady-state free precession (bSSFP) sequence, which was triggered to mid-diastole during expiration breath-hold. B₀ inhomogeneity was inspected in a mid LV SA plane using a ¹H GRE (FOV = 400 mm × 400 mm, spatial resolution = 6.25 mm × 6.25 mm, 2 echoes with ΔTE = 3.30 ms, TE of 1^st echo = 2.80 ms) with breath-hold (no cardiac triggering). The SA plane and the shim currents were retained for ¹³C imaging. The HP ¹³C images were acquired with ECG gating. The scanning sequence began after a 25-s delay from the start of pyruvate injection using an automated breath-hold voice instruction given immediately before the scan. Subjects were instructed to hold their breath in expiration for ~20 s, followed by a shallow breathing to minimize the respiratory motion. The center frequency was set to the in vivo HP [¹³C]bicarbonate resonance, which was predetermined by the in vivo HP [1-¹³C]lactate frequency, and the center frequency of water protons was used to calculate the frequency of [1-¹³C]lactate with an empirically-derived scaling factor of 0.2514949. The multi-echo images of [¹³C]bicarbonate, [1-¹³C]lactate and [1-¹³C]pyruvate were acquired in an interleaved manner every 2 R-R intervals (single shot, FOV = 400 mm × 400 mm, spatial resolution = 16 mm × 16 mm, ST = 30 mm, #echo = 6, TE of 1^st echo = 16.99 ms, ΔTE = 9.59 ms, readout length = 57.53 ms/arm, receive spectral bandwidth = 104.4 Hz, FA = 60° for bicarbonate, 60° for lactate, and 20° for pyruvate, #timepoint = 16). The imaging protocol was approved by the local Institutional Review Board (IRB#: STU-2018–0227).

Table 1.

Subject characteristics and compartmentalized time-averaged T₂*s of HP [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate.

Subject Demographics						Time-Averaged T2* [ms]
Subject ID	Sex	Age [years]	Height [cm]	Heart Rate [bpm]	Weight [kg]	[1-13C]Pyr		[1-13C]Lac			[13C]Bic
						LV	RV	LV	RV	Myo	Myo
#1	F	38	172.7	72	55.0	83.3	133.0	47.9	48.2	46.2	61.5
#2	M	45	188.0	60	83.0	126.7	136.0	43.1	38.0	51.2	66.1
#3	F	48	152.4	64	61.1	116.1	119.3	31.7	46.5	33.7	65.5
Mean ± Standard Deviation						108.7 ± 22.6	129.4 ± 8.9	40.9 ± 8.3	44.2 ± 5.5	43.7 ± 9.0	64.4 ± 2.5

Data Reconstruction and Statistical Analysis

All the acquired HP ¹³C data were reconstructed and analyzed using MATLAB. Images from different echoes were reconstructed by inverse Fourier transform after gridding to Cartesian coordinates, apodization with a Gaussian function, sample density compensation, and a four-fold zero-filling in k-space. To combine the multi-echo data, images from all echoes were added with equal weights.

Off-resonance correction for the HP ¹³C images was conducted with a frequency-segmented method as described in.¹⁹ The data was demodulated over a range of off-resonance frequencies from −50 Hz to 50 Hz in 5 Hz step. From each of the resultant images, pixels whose field map values most closely match the reconstruction frequency were selected to form the final image. The ¹³C field map was estimated from the ¹H field map, which firstly passed a median filter (3 × 3) to eliminate noise spikes and was then resampled to the resolution of ¹³C image and scaled by the ratio of $\frac{{\mathit{\gamma }}_{{13}_{\mathit{C}}}}{{\mathit{\gamma }}_{1_{\mathit{H}}}}$.

For the animal data, regions of interest (ROIs) that contain hyperintense ¹³C signals were drawn in the ¹³C images, and the mean ¹³C signal within each ROI was fitted to an exponential decaying function along the echo times to calculate the T₂*s of HP [¹³C]bicarbonate, [1-¹³C]lactate and [1-¹³C]pyruvate. Paired Student’s t-test (α = 0.05, two-tailed) was performed between the consecutively measured T₂*s of HP metabolites to verify the reproducibility of the proposed method. Dynamic change of the T₂* measurements along the timepoints was evaluated by the coefficient of variation (CV).

For the human ¹³C data, LV, RV, and Myo were manually segmented based on the ¹H SA MRI, as shown in Fig. 7(A). The spatially averaged ¹³C signals over the compartments at each echo were used to calculate the compartmentalized T₂*s in the heart. To ensure the accuracy of T₂* measurement, signals from the first m echoes (m = 3 or 4, depending on the SNR) were used for linear regression with the premise that coefficients of determination (r²) for the linear regressions of first m and m+1 echoes are both larger than 0.95.

Figure 7. Hyperpolarized ¹³C imaging from a healthy human subject.

(A) The prescribed SA plane for ¹³C imaging (shown in ¹H MRI). (B) Δf₀ map of the corresponding slice. (C) Dynamic ¹³C images (combination of 6 echoes) of HP [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate, acquired after an injection of HP [1-¹³C]pyruvate.

To evaluate the improvement of SNR with the sequence, peak SNRs from the first-echo and echo-combined HP ¹³C images were calculated by:

$${\mathit{S}\mathit{N}\mathit{R}}_{\mathit{p}\mathit{e}\mathit{a}\mathit{k}}=\frac{{\mathit{s}}_{\mathit{m}\mathit{a}\mathit{x}}-\mathit{m}\mathit{e}\mathit{a}\mathit{n}\left(\mathit{n}\right)}{\mathit{s}\mathit{t}\mathit{d}\left(\mathit{n}\right)}$$ #(2)

where s_max is the maximum signal of the image and n is the background noise. For human images, SNR maps were also reported for both first-echo and echo-combined images.

Results

¹H MRI and ¹³C images of the [¹³C]bicarbonate phantom are shown in Fig. 3(A). The T₂* calculated from the multi-echo images was 54.5 ms (r² = 0.99), Fig. 3(B). T₂* of the phantom was estimated as 54.0 ms using the Lorentzian fitting (red dotted line) of the acquired spectrum (blue solid line) in magnitude as in Fig. 3(C).

Figure 3. Phantom evaluation.

(A) Performance evaluation of the pulse sequence using a [¹³C]bicarbonate phantom. The T₂* from the MESI sequence (B) was compared to that from the Lorentzian fit of ¹³C spectrum (C).

Time-resolved (15 – 40 s from the start of injection) cardiac ¹³C images of HP [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate from a representative rat are shown in Fig. 4. The echo-combined ¹³C images are overlaid on top of the corresponding ¹H image (Fig. 4(A)). ¹H Δf₀ map of the imaging slice is shown in Fig. 4(B). Strong pyruvate and lactate signals were detected from the LV, and a large bicarbonate signal was observed in the left anterior Myo, as previously reported.^9,10 The cardiac T₂* of HP [1-¹³C]pyruvate was measured as 24.9 ± 5.0 ms. T₂*s of [1-¹³C]lactate and [¹³C]bicarbonate were 16.4 ± 4.7 and 16.9 ± 3.4 ms, respectively.

Figure 4. Hyperpolarized ¹³C imaging from a representative rat.

An axial slice that includes the heart was prescribed for both ¹H and ¹³C MRI. (A) T₂-weighted image and (B) Δf₀ map were acquired in ¹H. (C) Multi-echo images (10 echoes) of [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate were acquired every 5 s after an injection of HP [1-¹³C]pyruvate, and combined at each timepoint.

Individual echo images could be also resolved at each timepoint. The first six echo images at the second timepoint (20 s post-injection), for instance, are shown in Fig. 5(A) from the representative rat. Changes of spatially-averaged ¹³C signals along the echo time are displayed in Fig. 5(B) for selected ROIs (solid black squares in Fig. 5(A)). For [1-¹³C]pyruvate and [1-¹³C]lactate, the mean ¹³C signals in the first four echoes were used for linear regression. For HP [¹³C]bicarbonate, the first three echoes were used due to the limited SNR in the images with longer echoes. The T₂*s of [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate were 27.3 ms (r² = 0.99), 9.5 ms (r² = 0.99) and 13.9 ms (r² = 0.99), respectively, at 20 s post-injection (Fig. 5B) from the representative rat.

Figure 5. In vivo cardiac T₂* measurement from the representative rat.

(A) The first six echo images of [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate at 20 s. (B) Signal decay of HP ¹³C metabolites in the ROI (black squares in (C)) along echo time. (C) T₂*s of HP signals at different timepoints. The reference lines denote mean ± standard deviation.

T₂*s were consistent over time (CV < 0.23). From the representative rat, the time-averaged T₂*s of [1-¹³C]pyruvate, [1-¹³C]lactate, and [¹³C]bicarbonate were measured as 25.0 ± 3.5, 12.9 ± 2.2, and 14.8 ± 1.1 ms, respectively, as shown in Fig. 5(C). No significant difference was found from the two consecutive measurements of time-averaged T₂*s, as summarized in Fig. 6.

Figure 6. Reproducibility of T₂* measurements in rats.

T₂*s of the HP ¹³C-metabolites measured from two consecutive HP [1-¹³C]pyruvate injections were consistent (p > 0.05). Each marker represents a T₂* measurement from a single timepoint and the error bars denote the mean ± standard deviation.

For HP ¹³C cardiac imaging of healthy subjects, pyruvate and lactate signals were primarily detected in LV and RV while bicarbonate was preserved in Myo, which is consistent with a previous report.⁵ T₂* of HP [1-¹³C]pyruvate was 108.7 ± 22.6 ms in LV and 129.4 ± 8.9 ms in RV. T₂* of [1-¹³C]pyruvate in Myo was not calculated due to the low SNR. [1-¹³C]Lactate T₂* was measured as 40.9 ± 8.3 ms in LV, 44.2 ± 5.5 ms in RV, and 43.7 ± 9.0 ms in Myo. T₂* of [¹³C]bicarbonate was reliable only in Myo and measured as 64.4 ± 2.5 ms, which is within the range of previous measurement from human heart (T₂* = 58 ± 20 ms) using the spectral linewidth of HP [¹³C]bicarbonate.¹³ Time-averaged T₂*s from individual subjects are summarized in Table 1. Fig. 7 shows HP ¹³C metabolite maps from a representative subject (subject 1). As shown in Fig. 7(C), HP [1-¹³C]pyruvate, [1-¹³C]lactate, and [¹³C]bicarbonate images were acquired at multiple timepoints, and multi-echo images at each timepoint were examined for T₂* estimation (Fig. 8). From the subject, T₂*s of [1-¹³C]pyruvate and [1-¹³C]lactate were 85.7 ms (r² = 0.99) and 56.6 ms (r² = 0.98), respectively, in LV at 25 s post-injection. In RV, the T₂*s of [1-¹³C]pyruvate and [1-¹³C]lactate were 164.2 ms (r² = 0.98) and 42.9 ms (r² = 0.99), respectively. T₂*s of [1-¹³C]lactate and [¹³C]bicarbonate in Myo was 49.2 ms (r² = 0.95) and 63.6 ms(r² = 0.97), respectively. T₂*s were consistent over the timepoints for all the HP metabolites (CV < 0.21), as shown in Fig. 8(C). The QC measured that the pyruvate concentration in the final solution was 248.3 ± 11.2 mM with 0.4 ± 0.3 μM residual radical. The liquid-state polarization level was measured as 33.5% ± 9.6%. The dissolution to injection time was 64.7 ± 3.8 s.

Figure 8. In vivo cardiac T₂* measurement from the representative human subject.

(A) Multiple echo images of HP [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate at 25 s post-injection. (B) Signal decays of HP signals in LV, RV, and Myo (ROIs are depicted as red contours in Fig. 7 (A)) along the echo time. (C) Dynamic changes of the T₂*s in the ROIs. The reference lines denote mean ± standard deviation.

Although T₂*-weighted, multi-echo images of HP metabolite maps can be combined to improve the SNR, the improvements varied for different metabolites depending on the T₂*s and the initial SNRs of the first-echo images. In animals, the peak SNRs of the echo-combined ¹³C images increased by 44 ± 22 %, 29 ± 23 %, and 28 ± 22 % compared to those of the first-echo images for [¹³C]bicarbonate, [1-¹³C]lactate and [1-¹³C]pyruvate, respectively. In humans, the peak SNRs increased by 135 ± 22 % for [¹³C]bicarbonate, 42 ± 18 % for [1-¹³C]lactate, and 105 ± 33 % for [1-¹³C]pyruvate. The echo-combined SNR improvements are summarized in Supporting Information Fig. S1.

Discussion

In this study, we implemented ECG-gated metabolite-selective ¹³C MESI sequence for cardiac T₂* mapping of HP metabolites. The feasibility and the reproducibility of T₂* measurement using the proposed method was validated in a phantom and animals. In humans, T₂*s of HP ¹³C-metabolites could be spatially resolved for LV, RV and Myo despite the limited spatial resolution due to the short echo-spacing and the compromised gradient performance of the wide-bore system.

Contributing Factors to T₂*s of HP ¹³C Metabolites

The T₂* values were reported in the heart for three HP metabolites: [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate. The distinct T₂* values could originate from several contributing factors such as off-resonance, blood flow, and T₂, in addition to acquisition parameters. The off-resonance could be from the spatial inhomogeneity of B₀ or inaccurate assignment of acquisition frequencies for each metabolite. Comparing the T₂*s measured from rat and human hearts, the longer T₂*s of HP metabolites in human heart may arise from the more homogeneous B₀ field (e.g., better shimming) due to the slower heart rate and the presence of ECG gating. (R2.9) Likewise, the inter-subject variance of B₀ inhomogeneities may explain the relatively wide distribution of T₂*s in rats, as shown in Fig. 6. Pyruvate had significantly longer T₂*s than lactate (p < 0.0001) and bicarbonate (p < 0.001); this is likely due to the T₂ difference between HP metabolites. Indeed, a previous rat study reported longer T₂ for [1-¹³C]pyruvate compared to [1-¹³C]lactate in the liver.²⁰ Moreover, it was noted that T₂*s of HP [1-¹³C]pyruvate and [1-¹³C]lactate were longer in RV compared to those in LV (Table 1). The difference could be related to the higher filling velocity in LV than RV.²¹ Flow-induced signal decay is strong in LV, resulting in shorter T₂*s for HP [1-¹³C]pyruvate and [1-¹³C]lactate in LV than RV. On the contrary, the relatively long T₂* of HP [¹³C]bicarbonate in Myo is likely due to the absence of the blood flow effect.

Improved Assessment of Cardiac HP Signals

For most HP studies, utility of measured HP T₂*s lies primarily in accurate quantification of HP signals. In particular, understanding T₂* effect is crucial for appropriate assessment of HP data acquired by GRE-based pulse sequences and the design of the readout trajectories. Previous efforts to compensate the T₂* effect in HP ¹³C images have been focused on the off-resonance effect. Reed, et al. demonstrated improved B₀ shimming to mitigate the non-uniform circumferential HP [¹³C]bicarbonate signal from the Myo.¹³ Traechtler, et al. proposed a model-based reconstruction to correct the B₀-induced phase offsets in HP ¹³C multi-echo acquisition.¹² Considering the other contributing factors besides off-resonance effect, direct measurements of in vivo T₂* would further benefit the cardiac HP ¹³C studies.

Potential Improvements and Applications

In this study, in-plane spatial resolution of in vivo HP studies was compromised due to SNR available and the echo time needed for accurate T₂* measurement. The partial volume effect may cause mismatch between the segmentation from ¹H MRI and actual ¹³C signal distribution, leading to inaccurate measurement of T₂*. The low in-plane resolution could also result in sub-optimal T₂* maps. Therefore, in this study, we took the spatially averaged signal for T₂* fitting instead of the voxel-by-voxel analysis to insure the accurate measurement of T₂*. For future work, multi-shot spiral readout can be considered with carefully designed excitation angle strategy.²² Alternatively, acceleration techniques such as parallel imaging²³ and compressed sensing²⁴ can also be integrated into the MESI sequence for improving spatial resolution.

In the rat studies, cardiac gating was not available for HP imaging, resulting in motion artifacts in the ¹³C images. By carefully choosing ROIs for the mean signal quantification, we were able to mitigate the impact from the artifact, while it would be beneficial to include cardiac gating or respiratory control for animal study in the future.

Several aspects of T₂*s in HP metabolites need further investigation. For instance, flow contribution to the T₂* can be clarified by utilizing flow-sensitive gradients.²⁵ Moreover, it should be noted that the T₂*s depend on acquisition parameters such as the size^26,27 and the shape²⁷ of the voxel, the slice thickness,²⁸ eddy currents.²⁹

As HP pyruvate is utilized to study metabolism in various in vivo applications,^30–32 T₂*s of HP ¹³C metabolites in other organs and tissue types will be beneficial for establishing optimized k-space sampling patterns. In particular, large variation of magnetic susceptibility is expected in pathological states such as glioblastoma³³ due to blood depositions and calcifications, leading to potential underestimation of HP signals.

In vivo T₂ is another under-explored feature of HP molecules.^20,34,35 Yen, et al. reported nearly 1-second-long T₂s of HP signals with potentially unique metabolic contrast in rat hepatocellular carcinoma.²⁰ Joe, et al. estimated in vivo T₂s of HP ¹³C-metabolites by combining the ¹H B₀ map and ¹³C T₂* maps.³⁵ Due to the complex T₂* relaxation mechanism of HP ¹³C-metabolites, T₂ calculation using ¹H B₀ map will need further verification.

Conclusions

In conclusion, we proposed a ¹³C multi-echo spiral imaging sequence to measure in vivo cardiac T₂*s of HP [1-¹³C]pyruvate and its products. A phantom test and in vivo animal studies validated the method is robust and reproducible. Cardiac T₂*s of HP [1-¹³C]pyruvate, [1-¹³C]lactate and [¹³C]bicarbonate at 8 mm × 8 mm spatial resolution in rodents were measured as 24.9 ± 5.0, 16.4 ± 4.7, and 16.9 ± 3.4 ms, respectively. T₂*s of the HP ¹³C-metabolites at 16 mm × 16 mm spatial resolution in human heart are observed as 108.7 ± 22.6 ms (LV) and 129.4 ± 8.9 ms (RV) for [1-¹³C]pyruvate, 40.9 ± 8.3 ms (LV), 44.2 ± 5.5 ms (RV) and 43.7 ± 9.0 ms (Myo) for [1-¹³C]lactate, and 64.4 ± 2.5 ms (Myo) for [¹³C]bicarbonate.