Next Article in Journal / Special Issue
An Online Repository for Pre-Clinical Imaging Protocols (PIPs)
Previous Article in Journal
Textural Features of Mouse Glioma Models Measured by Dynamic Contrast-Enhanced MR Images with 3D Isotropic Resolution
Previous Article in Special Issue
Animal Models and Their Role in Imaging-Assisted Co-Clinical Trials
Peer-Review Record

Metabolite-Specific Echo Planar Imaging for Preclinical Studies with Hyperpolarized 13C-Pyruvate MRI

Tomography 2023, 9(2), 736-749;
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Tomography 2023, 9(2), 736-749;
Received: 28 January 2023 / Revised: 17 March 2023 / Accepted: 19 March 2023 / Published: 27 March 2023

Round 1

Reviewer 1 Report

The authors describe the implementation of a spectroscopically-selective EPI sequence for imaging hyperpolarized pyruvate and lactate on a preclinical Bruker 3T MRI.  While these developments largely parallel earlier efforts on clinical imagers, the results are nevertheless of interest as they enable a more direct correlation between clinical and pre-clinical data, thereby providing a potentially valuable tool for so-called ‘co-clinical’ trials.


The authors present comparisons between their spatial-spectral EPI sequence and simpler CSI approaches, and assess the performance of the EPI approach as a function of spatial and temporal resolution and the choice of tip angle.


Overall, these results are likely to be of interest to preclinical imaging researchers, particularly if the pulse sequences can be disseminated in a form that enables translation of these methods to other institutions.  Some aspects of the description are vague and further explanation would be helpful.  Certain conclusions appear to be based on very limited in vivo data, and this should be acknowledged as a limitation or further data should be added.


Specific issues that should be addressed:


1.     In the introduction, other approaches such as bSSFP and EPSI should be referenced and briefly discussed.  For bSSFP approaches, see, e.g. Shang et al, (Magn Reson Med. 2017 Sep; 78(3): 963–975.) and earlier references therein.   Some of these references include preclinical imaging on high-field systems.

2.     On line 73, please include further details about the input function used in the simulations.

3.     In the discussion of Fig. 1, the spatial spectral pulse has a passband of 120Hz (~4ppm at 3T).  Please comment on whether spectral selective imaging of the relatively small alanine resonance (6ppm from pyruvate) would be practical with this approach for realistic off-resonance conditions in the body.  How much pyruvate ‘contamination’ might be expected, for instance, in the liver as a result of off-resonance excitation of the much larger pyruvate resonance?

4.     Throughout the paper, more details are needed on the CSI acquisition.  Please include some details on the spectral analysis.  Also, the CSI tip angle may be too large for optimal C13 imaging.  Nominally, an 8x8 CSI acquisition at 10 degree tip angle requires 64 excitations and consumes roughly two-thirds of the magnetization.  Repeated imaging at 4.25s temporal resolution may deplete pyruvate magnetization so rapidly that the lactate signal never has a chance to accumulate.  A brief discussion of the choice of tip angle, and how this tip angle may affect the subsequent results—especially the PSF and the lactate signal, should be included where appropriate (e.g. Figs. 2, 3 and 6).

5.     In the PSF plots of in Fig. 2, please plot both the CSI PSF as well as the standard ‘sinc’ PSF of the EPI acquisition to enable comparison.  Please also include the CSI PSF for a 5 degree tip angle.  [This was the tip angle chosen for the EPSI acquisition of Chen et al, MRM 2007 (]

6.     In the comparisons of Figures 3, 5, and 6 a number of datasets appear to have been acquired in the same animal during a single scan session.  Please include the delay between the two C13 scans.  For EPI/CSI comparisons, please include the order of imaging, as the large pyruvate dose of the first can can potentially impact subsequent imaging. 

7.     Fig. 5 is apparently based on imaging of single animals for each column, and with a different kidney model in the first column.  Also, in the two center plots, and the plot at lower right, it appears that the lactate and/or pyruvate SNR is quite limited for the 2mm acquisition (the lactate and pyruvate SNR appear to be consistent with 1 over the time course show in the upper center figure.)  The small number of animals and the limited SNR impact the ability to draw strong conclusions, and this should be noted in the discussion of lines 219-229.

8.     Lastly, the great promise of this methodology lies in its eventual application to co-clinical imaging.  It would be highly illuminating, wherever possible, to highlight the similarities and differences seen in the preclinical data relative to clinical data.  For instance, how do the different timescales for vascular delivery seen in patients and preclinical models impact the pulse sequence optimization?

Author Response

  1. We added discussion and references to bSSFP and EPSI in the introduction (L50-52,L63-66)
  2. We added more details about input function (L90-91): “A realistic input function was modeled as a gamma distribution with FWHM of 8s determined from previous in vivo datasets. The input function was used to simulate bolus effects and inflow of injected pyruvate.”
  3. Yes, it would be possible to acquire alanine with the given spectral spatial pulse. This pulse has been previously used to acquire alanine signal clinically (For example, see Lee et al. J Magn Reson Imaging. 2022; 56(6):1792-1806. PMCID: PMC9562149). Looking at this quantitatively, using our measured spectral response (Figure 1B) we would only expect about 1-1.5% of the pyruvate signal ‘contaminating’ alanine.
  4. We added more info on spectral analysis and a citation (L165). In a previous paper, we used 10 degrees for CSI and found it worked well (see Also, we added example of CSI dynamics (See Figure S1) to show that the lactate dynamics do not decay quickly.
  5. Figure 2 was adjusted to include PSF profiles for EPI acquisition as well and CSI plots for a 5 degree flip angle.
  6. We added to the methods (L152-154): “The average time between hyperpolarized imaging experiments was 41 minutes. For 8 studies where both spspEPI and CSI imaging was acquired, CSI was acquired first for 6 of them before spspEPI experiments.” Also, we added to figure 3 caption “For both examples, CSI was acquired before EPI.”
  7. We added the following language in the results section (L304-307): “Nevertheless, it’s important to consider that these comparisons were conducted with a very limited sample size and some of the data had limited SNR (Figure 5, center top) which interferes with drawing strong conclusions. Instead, these results prove to support previously shown simulations (Figure 4).”
  8. Vascular delivery shouldn’t be a concern as we adjust the time between start of injection and imaging both preclinically and clinically. In this work, the 10s we use for this time is standard and allows us to capture peak signal. As long as the peak signal is acquired, there shouldn’t be other concerns regarding pulse sequence optimization. Other differences moving from preclinical to clinical studies may include setups with multiarray coils or larger field-of-view sizes.

Reviewer 2 Report

This study implemented a 2D metabolite-specific echo-planar imaging (EPI) sequences with spectral-spatial (spsp) excitation in preclinical hyperpolarized [1-13C] pyruvate studies. Various comparisons (e.g., partial volume effects, pyruvate and lactate SNR, and KPL map) were made between this spsp EPI and chemical shift imaging (CSI) on in vivo mice experiments. The results showed advantages of this spsp EPI sequence than the CSI method when measure hyperpolarized 13C-pyrucate in preclinical study. The structure and content of the manuscript is good, however, it needs to describe some specific concepts better, especially in the methods part.


Major concerns:


Q1, L121: Instead of “adding delay time” to adjust the temporal resolution, can you adjust the dwell time for your acquisition? In this case, with longer dwell time (lower receiver bandwidth), the SNR should get improved. Longer delay and lower temporal resolution may not be a disadvantage anymore. If your sequence/protocol has the feature to adjust bandwidth, extra experiments with longer dwell time is good to try, especially for the case when Temporal res = 4s.


Q2. L70&L73: Is that possible to show a bit more details about the “pharmacokinetic model”? For example, I would like to know more about how to account for the T1 decay for pyruvate and lactate during this metabolism process. Also this is important to understand the later part about the PSF simulation.

Same for the “realistic input function”, it’s probably not hurt to show the mathematical equations to help the audiences understand this simulation better.



Minor concerns:


Q3. L41-L42: The sentence with “most popular” and “best performance” is a bit too extreme, I would suggest switch those to modest adjectives.


 Q4. L63-L64: “…optimize the different acquisition parameters signal acquisition…” the sentence cannot be read smoothly and it’s a bit difficult to understand. Rewrite this part.


Q5.L86: Show the sequence diagram for one TR. Including both the excitation and acquisition diagram for the pyruvate and lactate. Notations about how you adjusted the time delay for different temporal resolutions would also be helpful.


Q6. L109: How many mice has been scanned? And how did you group them with different xenograft tissues?


Q7. Figure 3: It would be nice to add arrows or markups to show the audiences where’re the tumor and the vasculature.

Q8. L209-L211: Although the simulation results have been listed in front of this part, it’s still a bit confusing to understand how the optimum flip angles and temporal resolution has been picked. Is there any quantitative analysis has been done? Or it’s just been decided qualitatively based on the observations/summary of the simulation.

Author Response

  1. Currently, to adjust temporal resolution we add a delay time after the acquisition of all the metabolites for one timepoint (see figure 1D for diagram). Adjusting the dwell time instead to adjust the temporal resolution wouldn’t be in our favor for a couple reasons. One, for thermal signals increasing the dwell time could increase the SNR, however for hyperpolarized signals that we are dealing with in this study, the relative increase of SNR we would get from increasing the dwell time would be negligible in comparison to the increased SNR due to dynamic nuclear polarization (DNP). Additionally, adjusting the dwell time could lead to more T2* decay as we currently choose our EPI acquisition time to get close to T2* values we expect (around 20-30ms). Finally, the delay added is currently there to make comparisons with the CSI sequence but in the future this time could be utilized to acquire multiple slices instead of having the delay.
  2. Equations were added to explain the pharmacokinetic model in greater detail (L73-108)
  3. Updated to “Metabolite-specific imaging is a popular tool in clinical hyperpolarized [1- 13C]pyruvate MRI as it provides excellent performance…”
  4. Updated to “In this, we include comparisons to CSI sequences, and additionally analyze different acquisition parameters as well as perform rate constant fitting for metabolite-specific EPI in murine studies”
  5. We added part D to figure 1 with a sequence diagram for one TR and across metabolites. Also, we added notations about delays to adjust temporal resolution.
  6. Details were added to the methods (L116-119): “Eight mice implanted with patient-derived xenograft tissues of either renal cell carcinoma (RCC), or prostate cancer (LTL610, LuCap93) in the kidney or liver were used for testing the sequences (RCC kidney: 3, LTL liver: 2, LTL kidney: 2, LuCap93 kidney: 1)[18–21].”
  7. We updated Figure 3 with vasculature and tumor notations.
  8. The simulation results (Figure 4) are used as a framework to guide parameter choices and the feasibility of these parameter choices are confirmed with the in vivo We did not necessarily use the strict maximum SNR or minimum kPL error to choose parameters but used that as a starting point. We added a paragraph to 3.3.2 (L286-293) to explain how we arrived at flip angle and temporal resolution choices.


Back to TopTop