Repeatability of Quantitative Magnetic Resonance Imaging Biomarkers in the Tibia Bone Marrow of a Murine Myelofibrosis Model

Round 1

Reviewer 1 Report

This is a fine MS describing repeatability statistics of QMRI in a mouse model of myelofibrosis. This is very impressive work and I do not have any substantial comments to further improve it. Given the author's precision and their expertise, I wonder if they could enhance the discussion by commenting on  other time series analysis to test for trends that might be more complicated to implement but provide earlier detection of differences? This might broaden the perspective somewhat. Overall, a terrific contribution. 

Author Response

Reviewer 1:

This is a fine MS describing repeatability statistics of QMRI in a mouse model of myelofibrosis. This is very impressive work and I do not have any substantial comments to further improve it. Given the author's precision and their expertise, I wonder if they could enhance the discussion by commenting on other time series analysis to test for trends that might be more complicated to implement but provide earlier detection of differences? This might broaden the perspective somewhat. Overall, a terrific contribution.

C1.R1: We thank the Reviewer for appreciation of this work.  We added clarifying statements in Discussion both for detection of early changes and for possibility of improved protocols for enhanced sensitivity.

Reviewer 2 Report

Thank you very much for allowing me to review this manuscript. The authors describe a workflow to measure significant changes in bone marrow imaging biomarkers in a murine myelofibrosis model, with the hopes of eventually translating to a clinical biomarker that may obviate bone marrow biopsy for evaluating myelofibrosis. The manuscript is well-written and describes the technique in detail. I appreciate the authors’ effort to develop non-invasive biomarkers for myelofibrosis, but some concerns remain. I propose some points that could be improved upon in a revised version below.

Abstract

I would suggest characterizing the three distinct tibial sections analyzed in more detail in the abstract.

Introduction

Lines 49-52: It seems unnecessary to me to cite the University where the study was conducted and the funding source in this part of the manuscript. Affiliations are listed along with the authors’ names, and funding is stated later in the appropriate section. I would remove this sentence.

Lines 77-82: One of the limitations of bone marrow biopsy described by the authors is the analysis of a single anatomic site. However, despite being able to scan the whole body using MRI, there is a focus on the tibial region in this study. It would be beneficial to understand why this region was chosen and why the authors assume that the repeatability coefficient obtained in the tibial area will hold in other anatomic sites.

Materials and Methods

Lines 96-99: A total of 37 scan pairs were collected from a sample of 15. Individual animals were tested up to 5 times. Please elaborate on the criteria used to select which animals were tested more than once. If the animals which presented lower repeatability coefficients were measured consistently more often than the others, they might skew the repeatability coefficient downward.

Lines 215-218: A single reader manually segmented the analyzed tibias to define volumes of interest. Since the primary outcome of this study is repeatability, it would be important to include measurements obtained by at least one other image analyst or some measure of VOI overlap (IoU, Dice coefficient).

Lines 239-246: Further detailing how many cases were eliminated for each kind of artifact (e.g., as a percentage of total cases) would allow readers to gauge better the robustness of the proposed workflow for identifying test-retest changes in this animal MF model.

Results

Nothing to add

Discussion

Lines 425-439: The study's limitations are not clearly stated. At the present form, relevant limitations include the use of a single reader to determine tibial VOI and isolated analysis of the tibia, disregarding other long or axial bones. Moreover, the MRI acquisition parameters were extremely optimal, with near absolute zero cooling of the RF coils to increase SNR. In clinical situations, this will be unlikely to hold. The authors could account for that here.

Conclusions

Nothing to add

Author Response

Reviewer 2:

The manuscript is well-written and describes the technique in detail. I appreciate the authors’ effort to develop non-invasive biomarkers for myelofibrosis, but some concerns remain. I propose some points that could be improved upon in a revised version below.

 Abstract

I would suggest characterizing the three distinct tibial sections analyzed in more detail in the abstract.

R2.C1: Added as suggested

Introduction

Lines 49-52: It seems unnecessary to me to cite the University where the study was conducted and the funding source in this part of the manuscript. Affiliations are listed along with the authors’ names, and funding is stated later in the appropriate section. I would remove this sentence.

R2.C2: The text was modified to remove mention of the study site and keep the general reference to the CIRP program.

Lines 77-82: One of the limitations of bone marrow biopsy described by the authors is the analysis of a single anatomic site. However, despite being able to scan the whole body using MRI, there is a focus on the tibial region in this study. It would be beneficial to understand why this region was chosen and why the authors assume that the repeatability coefficient obtained in the tibial area will hold in other anatomic sites.

R2.C3: The text was modified to clarify that tibia was chosen to test for bone marrow inhomogeneity without motion artifacts and to fit into the confined space of the cryogenic imaging coil for enhanced SNR and resolution. As described in Methods, this is an induced, transplant model of MF disease in mice post-ablation, which is different from human whole-body application.

Materials and Methods

Lines 96-99: A total of 37 scan pairs were collected from a sample of 15. Individual animals were tested up to 5 times. Please elaborate on the criteria used to select which animals were tested more than once. If the animals which presented lower repeatability coefficients were measured consistently more often than the others, they might skew the repeatability coefficient downward.

R2.C4: Clarified that TT-RT sessions were performed for majority of longitudinal scans prior to evaluation of image quality.  RC was evaluated from all available TT-RT pairs after completion of the experiment (rather than for individual mice).

Lines 215-218: A single reader manually segmented the analyzed tibias to define volumes of interest. Since the primary outcome of this study is repeatability, it would be important to include measurements obtained by at least one other image analyst or some measure of VOI overlap (IoU, Dice coefficient).

R2.C5: We agree with Reviewer’s comment on importance of repeatable segmentation. In fact, multiple factors contribute to the measured RC, including the animal model, consistent repositioning, MRI system stability, and segmentation. This work focused on establishing the overall repeatability of mean MRI metrics without isolated analysis of the individual contributing variables.  We have now stated explicitly that inter-observer repeatability was not assessed in our study and added this as a potential error source in Discussion.

Lines 239-246: Further detailing how many cases were eliminated for each kind of artifact (e.g., as a percentage of total cases) would allow readers to gauge better the robustness of the proposed workflow for identifying test-retest changes in this animal MF model.

R2.C6: We added the requested details on excluded TT-RT pair fractions (included pairs are also summarized in Table 1).

Discussion

Lines 425-439: The study's limitations are not clearly stated. At the present form, relevant limitations include the use of a single reader to determine tibial VOI and isolated analysis of the tibia, disregarding other long or axial bones. Moreover, the MRI acquisition parameters were extremely optimal, with near absolute zero cooling of the RF coils to increase SNR. In clinical situations, this will be unlikely to hold. The authors could account for that here.

R2.C7: We added a paragraph on the study limitations, as suggested.