Using light and X-ray scattering to untangle complex neuronal orientations and validate diffusion MRI
Curation statements for this article:-
Curated by eLife
eLife assessment
This paper presents a valuable cross-validation study of mesoscopic measurements of axonal orientations from three different modalities: x-ray tomography, scattered light imaging, and diffusion MRI. The authors show convincing similarities and differences in fibre orientations from all three methods over partial ex vivo brain samples, though as only a single diffusion method is investigated, there is inadequate evidence to support conclusions about diffusion MRI reconstruction methods in general. As a first example of work comparing these three modalities, it is of interest to researchers who want to apply x-ray tomography or scattered light imaging to image the white matter ex vivo or use these methods for future validation of diffusion MRI methods.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Disentangling human brain connectivity requires an accurate description of nerve fiber trajectories, unveiled via detailed mapping of axonal orientations. However, this is challenging because axons can cross one another on a micrometer scale. Diffusion magnetic resonance imaging (dMRI) can be used to infer axonal connectivity because it is sensitive to axonal alignment, but it has limited spatial resolution and specificity. Scattered light imaging (SLI) and small-angle X-ray scattering (SAXS) reveal axonal orientations with microscopic resolution and high specificity, respectively. Here, we apply both scattering techniques on the same samples and cross-validate them, laying the groundwork for ground-truth axonal orientation imaging and validating dMRI. We evaluate brain regions that include unidirectional and crossing fibers in human and vervet monkey brain sections. SLI and SAXS quantitatively agree regarding in-plane fiber orientations including crossings, while dMRI agrees in the majority of voxels with small discrepancies. We further use SAXS and dMRI to confirm theoretical predictions regarding SLI determination of through-plane fiber orientations. Scattered light and X-ray imaging can provide quantitative micrometer 3D fiber orientations with high resolution and specificity, facilitating detailed investigations of complex fiber architecture in the animal and human brain.
Article activity feed
-
-
Author Response
Reviewer #1 (Public Review):
This study presents a valuable comparison of fibre orientation estimates from three different modalities: diffusion MRI, scattered light imaging, and x-ray scattering. The comparison is interesting as each modality is sensitive to different aspects of tissue microstructure - water anisotropy, micron-scale structural coherence, and myelin lamella respectively. Where scattered light and x-ray imaging can be only applied ex vivo, diffusion MRI has in vivo applications but suffers from being an indirect estimate of the microstructure of interest. By acquiring all modalities in both a vervet monkey and human brain sample, the authors provide quantitative, pixel/voxel-wise comparisons of fibre orientation estimates within the same tissue samples. The authors show convincing agreement in fibre …
Author Response
Reviewer #1 (Public Review):
This study presents a valuable comparison of fibre orientation estimates from three different modalities: diffusion MRI, scattered light imaging, and x-ray scattering. The comparison is interesting as each modality is sensitive to different aspects of tissue microstructure - water anisotropy, micron-scale structural coherence, and myelin lamella respectively. Where scattered light and x-ray imaging can be only applied ex vivo, diffusion MRI has in vivo applications but suffers from being an indirect estimate of the microstructure of interest. By acquiring all modalities in both a vervet monkey and human brain sample, the authors provide quantitative, pixel/voxel-wise comparisons of fibre orientation estimates within the same tissue samples. The authors show convincing agreement in fibre orientations from all three methods, giving confidence in the fidelity of the methods for neuroanatomical investigations. Differences are also observed: SLI is shown to have less reliable estimates of fibre inclination, and the CSD analysis presented overestimates the number of crossing fibre populations when compared to the microscopy methods, particularly in single fibre regions such as the corpus callosum, a known artefact in some diffusion analyses.
In the current PDF, it is very difficult to see fibre orientations in figures due to low resolution, limiting the reader's ability to assess the results. Higher-resolution images would provide more information and easier comparisons.
The methods are generally clear though some additional information is needed:
- to specify the resolution that the orientations are compared in each figure and how data was up-/down-sampled for these comparisons respectively. For example, each SAXS pixel contains many SLI pixels. It is currently unclear whether the mean SLI orientation from a neighbourhood is equivalent to the SLI compared, or whether a comparison was made for each SLI pixel. Similarly, for the dMRI-microscopy comparisons.
- I also could not follow why two SLI methods are presented in the methods: SLI scatterometry relating to Figure 2, and angular SLI relating to all other results. Further clarification is needed.
- Since the quality of the data co-registration can strongly impact pixel/voxel-wise comparisons, quantification of the registration accuracy or overlays demonstrating the quality of the co-registration would be valuable.
A primary weakness of the work as a diffusion MRI validation study is that though diffusion MRI supports many different models to extract fibre orientations with different outputs, here only a single model is compared to the microscopy data, which may affect the generalisability of the results. Further, it only compares the primary orientations from the diffusion MRI and does not consider each fibre population's magnitude (density of fibres) or the orientation dispersion, both of which can influence downstream analyses.
The paper could be strengthened by a more detailed discussion on the differences between the imaging modalities - e.g. in terms of imaging resolution, signal-generating mechanisms, and sensitivity to specific aspects of the tissue microstructure - and how these differences may limit their application to specific neuroanatomical investigations, or ability to validate one another. For example, the microscopy sections are 80 microns thick whilst the diffusion voxel is 200 microns. I expect this could contribute to the difference in the number of fibre populations per voxel.
The hypothesis that dMRI signal contributions from extra-axonal water result in additional fibre populations could be investigated by running CSD on both low and high-b-value data (for example using the openly available MGH dataset, Fan 2016) where fewer secondary fibre populations should be observed at high b-value.
We sincerely thank Reviewer #1 for the constructive feedback, which helped us to significantly improve our manuscript. We hope to have done our best to address all concerns:
First, we regret the insufficient resolution of figures. The resolution must have been reduced during the submission process, when generating the pdf version of our manuscript. We have now submitted all figures as separate files with the highest possible resolution. In addition, all parameter maps are publicly available and can be opened and zoomed in, e.g. with ImageJ, to see the fiber orientations of individual image pixels.
As requested by the reviewer, we have modified our manuscript and added additional methods information.
Concerning the data up-/downsampling: We have now specified in each figure caption at which resolution the images were compared and added the following explanation to the newly named Methods section “Image registration and pixel-wise comparison”: To minimize loss of information, the pixel/voxel-wise comparisons were performed at the spacing of the highest resolution image, i.e. the lower-resolution diffusion MRI (dMRI) and small-angle X-ray scattering (SAXS) images were upscaled to match the higher-resolution scattered light imaging (SLI) images. As a result, the fiber orientation of one SAXS pixel (px=150µm) was compared to the fiber orientations of 50x50 SLI pixels (px=3µm), and not to the mean; similarly for comparisons with dMRI.
Concerning the two SLI methods: We have added the following explanation to the Methods section “Scattered Light Imaging” to clarify why we used two different methods: To generate the scattering patterns (upper Figure 2C), a time-consuming SLI scatterometry measurement was performed in which the sample was illuminated from 6,400 different angles, as described in Menzel et al. (2021b). This was necessary to achieve sufficiently resolved scattering patterns for a visual comparison with SAXS scattering patterns. The fiber orientations can also be extracted from the peak positions in the azimuthal profiles (cf. bottom Figure 2C), without taking the overall shape of the scattering patterns into account. Therefore, all other results were obtained from more time- and data-efficient angular SLI measurements in which the sample was illuminated from 24 different angles around a circle and the fiber orientations were derived from the peak positions in the resulting line profiles, as described in Menzel et al. (2021a).
Concerning the quality of the co-registration: We thank the reviewer for this comment. We agree that the accuracy of image registration has a high impact on pixel/voxel-wise comparisons and determines the quality of our cross-validation study. We have added a new Discussion section “Quality of cross-validation” and inserted a new figure (Figure 4–figure supplement 1) to demonstrate the accuracy of image registration, both for the vervet and human brain samples: The reference and registered images are shown both in direct comparison (top and middle images, respectively) and as overlays (bottom images), as suggested by the reviewer. Reference and registered images show good correspondence (white/gray matter boundaries coincide). Only the fornix of the vervet brain section is not aligned (it moved when re-mounting the sample) so that this region was evaluated separately, as described in the manuscript. We found standard linear transformations (scaling, rotation, and translation) to be sufficient for achieving a fair comparison between the different modalities, demonstrating the experimental feasibility of our approach. There might still be individual voxels that were not sufficiently well aligned, especially when comparing sections (SLI/SAXS) to volumetric measurements (dMRI). However, this would only increase the angular differences between the fiber orientations. Our results can therefore be considered as an upper bound. Using standard linear transformations, we could already show that in-plane crossing orientations from SAXS and SLI, and through-plane orientations from SAXS and dMRI correspond very well to each other.
We understand the focus of our work lying rather on the cross-validation/evaluation of light and X-ray scattering, in comparison to dMRI which is much longer established, than on a “diffusion MRI validation study”: the myelin specific SAXS orientations and crossings were cross-validated with the high-resolution SLI orientations, and SLI out-of-plane fibers were validated using SAXS/dMRI as ground truth data.
The reviewer rightly noted that we used a single analysis method to extract fiber orientations from dMRI data (based on the MRtrix3 dwi2response and dwi2fod commands, using the dhollander and msmt_csd algorithms, respectively). Although to our knowledge this method is one of the most widely used for deriving fiber orientations for subsequent tractography, it is true that other methods might yield different results and that we cannot draw conclusions for diffusion MRI in general. We have included these considerations in the newly named Discussion section “Comparison of SAXS and SLI fiber orientations to dMRI”.
It is also true that our comparison focused on primary dMRI orientations without taking fiber density or dispersion into account. We decided to do so because deriving such metrics from SLI or SAXS data has not been implemented yet. However, we expect this to happen in the following years, enriching future studies. We have also included these aspects in the Discussion section.
We agree with the reviewer that our paper could be strengthened by a more detailed discussion on the differences between the imaging modalities. We have added a paragraph to the new Discussion section “Quality of cross-validation”: We compared results from three different imaging techniques (SLI, SAXS, dMRI) which all have different signal-generating mechanisms and resolutions. The different resolutions should be taken into account when interpreting the comparative studies. To investigate the relationship between SLI peak distance and fiber inclination, we used dMRI/SAXS images with at least 50 times lower in-plane resolution as reference (Figure 6). This is sufficient to validate the theoretical predictions, but insufficient to validate individual pixel values. To validate crossing fiber orientations from SAXS, we used SLI images with 30 times higher in-plane resolution, leading to a broad distribution of angular differences (depending on the region), but the mean difference around zero is evidence for a good overall correspondence (Figure 4). Finally, when comparing fiber orientations in SAXS and SLI to dMRI (Figure 5), it should be taken into account that dMRI voxels (with 200µm size) contain more fiber layers than the corresponding SAXS or SLI voxels (with 80µm section thickness), so that dMRI voxels might include additional fiber populations not present in SAXS or SLI data. On the other hand, fiber orientations that occur both in dMRI and SAXS voxels – like the out-of-plane fiber orientations from SAXS and dMRI (e.g. Figure 6B-C) – can be considered as reliable, given the substantially different contrast-generating mechanisms.
Finally, we thank the reviewer for the suggestion to study different b-values (last comment). We agree that an analysis based on different b-values might yield different results. Especially, an analysis with high b-values is expected to be more specific to the fiber orientations, as most other components of the signal would have already been attenuated. To investigate this hypothesis, we have run a separate analysis with high b-values only (5 and 10ms/μm2) and added a new supplementary figure (Figure 5–figure supplement 4) that compares the results for all b-values to high b-values only. We found that the fiber orientation distributions are almost identical between all b-values and high b-values only.
Reviewer #2 (Public Review):
This work is a cross-validation of an x-ray tomography technique (SAXS) and an optical microscopy technique (SLI) for imaging axonal orientations ex vivo. These innovative methods were introduced in recent papers by the authors, who have teamed up here to compare them side-by-side on the same tissue samples for the first time. The two methods are both label-free (do not require staining) and they are quite complementary. SAXS can provide full 3D orientation measurements on intact tissue, but it operates at a mesoscopic resolution and requires access to a synchrotron. SLI can measure the orientations of multiple fascicles per voxel at a microscopic resolution and relies on more widely accessible equipment, but its accuracy suffers for fiber orientations perpendicular to the imaging plane and it requires tissue to be sectioned before it is imaged. Therefore it makes a lot of sense to explore the complementary strengths of these two techniques, and to use one to "fill in the blanks" of the other. The paper also compares the orientation measurements obtained with SAXS and SLI to those obtained with diffusion MRI. The latter provides only indirect measurements based on water diffusion, at a mesoscopic resolution somewhat lower than that of SAXS, but has the benefit of being feasible in vivo.
A limitation of this study is that conclusions on the comparison between SAXS and SLI are drawn from only 2 sections of a partial monkey brain sample and 2 sections of a partial human brain sample. Conclusions on diffusion MRI are drawn only on the 2 human sample sections. This is particularly an issue for the comparison to diffusion MRI, as the diffusion MRI voxels are wider than the section thickness, hence one cannot preclude that any orientations detected with diffusion MRI but not with SAXS and SLI come from the portion of the voxel that is missing from the corresponding SAXS/SLI section.
The stated aim of the paper is to provide a framework for combining the complementary benefits of SAXS and SLI, rather than simply presenting the results of a cross-validation study. This is a significant and ambitious aim. However, in order for this to serve as a framework, there would have to be clear prescriptions for how researchers interested in obtaining ground-truth measurements of axonal orientations would do so by using these two methods in tandem. This is not adequately developed in the paper in its present form. For example, the results show reasonable agreement between SAXS and SLI orientations when fibers lie within the SLI imaging plane and decreasing agreement for fibers with increasing through-plane inclination. How would the two methods be combined in voxels where they disagree? Would one use SLI orientations in voxels with fewer through-plane fibers and SAXS orientations in voxels with more through-plane fibers? How would voxels be assigned to each category? How would the orientation vectors from the two modalities be composed and how would the resolution difference between the two be handled? When the through-plane measurement of SLI is unreliable, is its in-plane measurement still reliable? That is if there were one mainly in-plane and one mainly through-plane fiber population, would the orientation of the former still be measured correctly by SLI? There is also considerable agreement reported here between through-plane orientations obtained with SAXS and diffusion MRI. Would this mean that diffusion MRI itself could be used to supplement SLI with through-plane orientations? Any clear set of prescriptions along these lines would represent a framework for imaging orientations by combining modalities. This, however, would require detailed steps for how to perform the combination and use the multi- vs. uni-modal framework to reconstruct connectional anatomy.
A key advantage of SAXS is that it can be performed on intact samples, i.e., before any nonlinear distortions of the tissue are introduced by sectioning. Thus it can provide an undistorted reference, with contrast on axonal orientations that would be absent in, say, a structural MRI of comparable resolution. This contrast could be used to drive registration of the distorted SLI sections to an undistorted SAXS volume, and therefore is a key way in which the two techniques can complement each other. Here, however, this is not explored, as SAXS is performed after sectioning. It is not clear if this is the authors' prescription for how a combined SAXS/SLI framework would be implemented, or if it was done specifically for this study.
First, it would seem that SAXS on the intact sample would be lower maintenance, requiring less setup time and hence potentially less overall beamtime than performing SAXS on each section separately. This would make it more practical for routine deployment beyond a few sections.
Second, because the SAXS data are now nonlinearly distorted, they cannot be affinely aligned to the MRI volumes. While, in principle, performing both SAXS and SLI on the sections may facilitate the comparison between the two, having to unmount, rehydrate, and remount the sections in between may negate this advantage, as now there is no guarantee that SAXS and SLI can be affinely registered to each other. Here all these registration steps are performed affinely, so it is unclear to which extent the computed errors between modalities are characterizing the inherent limitations of the respective contrasts, or limitations of the registration technique. Some of the alignment is performed manually, for example, specific regions of the images are realigned by hand, and the slice of the diffusion MRI volume that is aligned to the SAXS/SLI sections is chosen by hand. Again, for this to serve as a framework that can be deployed on whole samples, there would have to be clear prescriptions for how to perform these steps robustly, how to ensure that the MRI can be acquired in a coordinate frame parallel to the sections, etc.
Finally, the paper puts forth a general conclusion that diffusion MRI overestimates the number of fiber populations per voxel, on the basis of small ODF peaks appearing perpendicular to the main ODF peaks. Of all conclusions in the paper, this is the least convincingly supported by evidence. First, these small perpendicular peaks are a known artifact, which would be typically eliminated by ignoring ODF peaks below a certain amplitude, a common practice in diffusion tractography algorithms. The authors refrain from using an amplitude threshold, with the rationale that it may also remove true diffusion orientations. However, they apply a threshold when they detect SLI peaks (a rather stringent 8% of the maximum). Second, the explanation that these artifactual peaks may appear due to vessel walls is not convincing. Vasculature is sparse. A single vessel wall will not impact the diffusion signal in the same way as a bundle of parallel axons. In an axon bundle, water molecule displacements are restricted in all directions except parallel to the axons. A single vessel wall in a voxel will not have the same effect on displacements (which are much smaller than the size of the voxel). From Figure 5, it looks like there would be at most 1-2 of these vessels in a diffusion MRI voxel, and they would not be in all voxels. This cannot explain the widespread appearance of these small artifactual peaks. Third, many ODF reconstruction methods have parameters that can be adjusted to make these artifactual peaks more or less prominent. The default parameters may be optimal for in vivo but not ex vivo data, due to the effects of fixation. In light of these concerns, I would caution against making such a general statement about all diffusion MRI in the human brain, especially on the basis of a single diffusion reconstruction method applied to a single location in one brain.
We sincerely thank Reviewer #2 for the constructive feedback, which helped us to significantly improve our manuscript. We hope to have done our best to address all concerns:
First, regarding the limited number of tissue sections used for our study (second paragraph):
It is true that we only evaluated a limited number of samples – mainly due to the limited beam time available for SAXS experiments. We believe that the main conclusions concerning the cross-validation of SAXS crossing fibers and SLI out-of-plane fibers still remain valid.
The reviewer correctly points out that the dMRI voxels (with 200um size) are wider than the section thickness (80um) so that additional fiber orientations detected with dMRI might come from the portion of voxels missing in the corresponding SAXS/SLI measurement. We have added a clarifying paragraph in the newly named Discussion section “Comparison of SAXS and SLI fiber orientations to dMRI” as well as in the new Discussion section “Quality of cross-validation”. Nevertheless, we do not expect additional fiber orientations in comparable homogeneous regions like the corpus callosum, and fiber orientations that occur both in dMRI and SAXS/SLI – like the out-of-plane fiber orientations from dMRI and SAXS (e.g. Figure 6B-C) – can be considered as reliable, given the substantially different contrast-mechanisms of the microscopy and dMRI techniques.
Concerning the aim of our paper and the questions raised by the reviewer in the third paragraph:
We understand that the term “framework” is not the appropriate word in this context, as it can raise false expectations. Our aim was rather to provide a basis (“groundwork”) to enable combined measurements of SLI/SAXS (and dMRI) on the same tissue samples and cross-validate the techniques (the crossing fiber orientations in SAXS and the through-plane fiber orientations in SLI have not been validated using other techniques so far). We have changed the wording throughout the manuscript, explaining that we focused on laying the “groundwork” instead of providing a “framework”, and reformulated the corresponding sentences.
Our aspiration was to provide a protocol how the complementary imaging techniques can be performed on the same tissue sample. When talking about a “combination” of techniques, we were referring to combined measurements (i.e. measurements on the same sample), and not to a combined analysis (e.g. in form of combined parameter maps and fiber orientation vectors). The latter, while very much needed in the field, would require many more and heterogeneous samples, and work beyond the scope of this manuscript, which we hope to perform in the future. Along these lines, we have removed the term “combined” throughout the manuscript, and wrote e.g. “measurements of SLI and 3D-sSAXS on the same tissue sample” instead of “combined measurements of SLI and 3D-sSAXS” to avoid confusion.
However, it is of course a valid question how SAXS and SLI can be combined in voxels where they disagree, how the orientation vectors can be composed, and how the resolution difference between the methods can be handled. We have added a new Discussion section “Towards a combination of SLI, SAXS, and dMRI” to elaborate on how a combined analysis (e.g. in form of combined fiber orientation maps) can be achieved and what challenges we are facing.
Concerning the reviewer’s question if the orientation of an in-plane fiber population would be correctly measured by SLI if there was another through-plane fiber population: We only evaluated regions belonging to a single fiber population (SLI azimuthal profiles with one or two dominant peaks) and regions belonging two in-plane crossing fiber populations (SLI azimuthal profiles with two dominant peak pairs). Voxels containing both in-plane and through-plane fibers were excluded from the analysis. The determined in-plane SLI orientations can thus be considered as reliable. We have added these aspects to the new Discussion section
“Quality of cross-validation”.
Regarding the reviewer’s question if dMRI itself could be used to supplement SLI with through-plane orientations: Diffusion MRI could indeed be used as a reference to enhance the interpretation of through-plane fiber orientations from SLI measurements. One disadvantage over SAXS is the lower resolution and that it cannot directly be performed on the same tissue section as SLI. These aspects have also been added to the new
Discussion section.
Concerning the reviewer’s suggestion to perform SAXS before sectioning and the problem of image registration (fourth paragraph):
It is true that SAXS tensor tomography can be applied to larger tissue volumes and that it is not limited to tissue sections. However, the reconstruction of crossing fibers has so far only been realized in sections (Georgiadis et al., 2022) and not in intact samples. As we wanted to cross-validate these fiber crossings using SLI as reference, we decided to perform the SAXS measurements on the same tissue sections as the SLI measurements. A comparison to results from SAXS tensor tomography might still be interesting in the future. We have added these considerations to the new Discussion section “Towards a combination of SLI, SAXS, and dMRI”.
It is also true that cutting a section from a brain tissue sample might introduce non-linear distortions; in particular, it is challenging to identify this particular section in the original tissue volume; unmounting and remounting of an already existing section introduces much less distortions. We have added a new figure (Figure 4–figure supplement 1) which shows that a co-registration with linear transformations (scaling, rotation, and translation) is already sufficient to allow for a fair comparison between the different image modalities, both for vervet and human brain samples. Only the fornix of the vervet brain section moved during remounting of the sample, and was therefore evaluated separately, as described in the manuscript. In any case, even if the angular differences in some image pixels were larger due to an imperfect co-registration, a perfect co-registration would only yield even smaller differences. Hence, the reported angular differences can be considered as upper bound, demonstrating that SAXS and SLI fiber orientations show already a very good correspondence. We have added a corresponding paragraph to the new Discussion section “Quality of cross-registration”.
Finally, we agree that a clear prescription would be necessary to enable combined analysis on whole tissue samples. As mentioned further above, our aim was to provide the groundwork for combined measurements on the same tissue sample and cross-validate the different techniques, and not to provide combined fiber orientation maps or similar. We have added our thoughts on how to combine the different image modalities to the new Discussion section “Towards a combination of SLI, SAXS, and dMRI”.
Concerning the final concern of the reviewer that an overestimation of the number of fiber populations per voxel is not sufficiently supported (last paragraph):
We understand this concern and have removed all phrases that could be understood as generalized claims for MRI, including any reference to fiber orientations overestimation. Furthermore, we have extended the Discussion to indicate the non-generalizability of our results.
Regarding the first point that the minor perpendicular ODF peaks could be removed by applying a suitable amplitude threshold: This is a valid remark and was discussed partly in the first version of the manuscript, when referring to increasing the threshold of secondary lobes prior to running tractography algorithms and to the problem that it might decrease the sensitivity for the cases where there exist actual but less prominent secondary fiber populations. We have extended the Discussion to address the concerns of the reviewer.
Regarding the second point that the minor ODF peaks are probably not caused by vessel walls: We thank the reviewer for the valid remarks and have removed all mentions of blood vessels in the manuscript, including the arrows in Figure 5H.
Regarding the third point that parameters can be adjusted to make the artifactual peaks more/less prominent, and that default parameters might be optimal for in vivo but not ex vivo data: We have added the remark that model parameters can be fine-tuned to decrease the percentage of false-positives to the Discussion.
Finally, it is true that we only used a single diffusion reconstruction method and measured only a single location in one human brain with dMRI. As mentioned at the very beginning, the number of samples was limited, and we included the reviewer’s concerns in the newly named Discussion section “Comparison of SAXS and SLI fiber orientations to dMRI”. For the main purposes of the paper like the cross-validation of out-of-plane fibers in SAXS/SLI, the dMRI data was still sufficient as we could show a good correspondence between dMRI/SAXS in these regions.
-
eLife assessment
This paper presents a valuable cross-validation study of mesoscopic measurements of axonal orientations from three different modalities: x-ray tomography, scattered light imaging, and diffusion MRI. The authors show convincing similarities and differences in fibre orientations from all three methods over partial ex vivo brain samples, though as only a single diffusion method is investigated, there is inadequate evidence to support conclusions about diffusion MRI reconstruction methods in general. As a first example of work comparing these three modalities, it is of interest to researchers who want to apply x-ray tomography or scattered light imaging to image the white matter ex vivo or use these methods for future validation of diffusion MRI methods.
-
Reviewer #1 (Public Review):
This study presents a valuable comparison of fibre orientation estimates from three different modalities: diffusion MRI, scattered light imaging, and x-ray scattering. The comparison is interesting as each modality is sensitive to different aspects of tissue microstructure - water anisotropy, micron-scale structural coherence, and myelin lamella respectively. Where scattered light and x-ray imaging can be only applied ex vivo, diffusion MRI has in vivo applications but suffers from being an indirect estimate of the microstructure of interest. By acquiring all modalities in both a vervet monkey and human brain sample, the authors provide quantitative, pixel/voxel-wise comparisons of fibre orientation estimates within the same tissue samples. The authors show convincing agreement in fibre orientations from all …
Reviewer #1 (Public Review):
This study presents a valuable comparison of fibre orientation estimates from three different modalities: diffusion MRI, scattered light imaging, and x-ray scattering. The comparison is interesting as each modality is sensitive to different aspects of tissue microstructure - water anisotropy, micron-scale structural coherence, and myelin lamella respectively. Where scattered light and x-ray imaging can be only applied ex vivo, diffusion MRI has in vivo applications but suffers from being an indirect estimate of the microstructure of interest. By acquiring all modalities in both a vervet monkey and human brain sample, the authors provide quantitative, pixel/voxel-wise comparisons of fibre orientation estimates within the same tissue samples. The authors show convincing agreement in fibre orientations from all three methods, giving confidence in the fidelity of the methods for neuroanatomical investigations. Differences are also observed: SLI is shown to have less reliable estimates of fibre inclination, and the CSD analysis presented overestimates the number of crossing fibre populations when compared to the microscopy methods, particularly in single fibre regions such as the corpus callosum, a known artefact in some diffusion analyses.
In the current PDF, it is very difficult to see fibre orientations in figures due to low resolution, limiting the reader's ability to assess the results. Higher-resolution images would provide more information and easier comparisons.
The methods are generally clear though some additional information is needed: 1) to specify the resolution that the orientations are compared in each figure and how data was up-/down-sampled for these comparisons respectively. For example, each SAXS pixel contains many SLI pixels. It is currently unclear whether the mean SLI orientation from a neighbourhood is equivalent to the SLI compared, or whether a comparison was made for each SLI pixel. Similarly, for the dMRI-microscopy comparisons. 2) I also could not follow why two SLI methods are presented in the methods: SLI scatterometry relating to Figure 2, and angular SLI relating to all other results. Further clarification is needed. 3) Since the quality of the data co-registration can strongly impact pixel/voxel-wise comparisons, quantification of the registration accuracy or overlays demonstrating the quality of the co-registration would be valuable.
A primary weakness of the work as a diffusion MRI validation study is that though diffusion MRI supports many different models to extract fibre orientations with different outputs, here only a single model is compared to the microscopy data, which may affect the generalisability of the results. Further, it only compares the primary orientations from the diffusion MRI and does not consider each fibre population's magnitude (density of fibres) or the orientation dispersion, both of which can influence downstream analyses.
The paper could be strengthened by a more detailed discussion on the differences between the imaging modalities - e.g. in terms of imaging resolution, signal-generating mechanisms, and sensitivity to specific aspects of the tissue microstructure - and how these differences may limit their application to specific neuroanatomical investigations, or ability to validate one another. For example, the microscopy sections are 80 microns thick whilst the diffusion voxel is 200 microns. I expect this could contribute to the difference in the number of fibre populations per voxel.
The hypothesis that dMRI signal contributions from extra-axonal water result in additional fibre populations could be investigated by running CSD on both low and high-b-value data (for example using the openly available MGH dataset, Fan 2016) where fewer secondary fibre populations should be observed at high b-value.
-
Reviewer #2 (Public Review):
This work is a cross-validation of an x-ray tomography technique (SAXS) and an optical microscopy technique (SLI) for imaging axonal orientations ex vivo. These innovative methods were introduced in recent papers by the authors, who have teamed up here to compare them side-by-side on the same tissue samples for the first time. The two methods are both label-free (do not require staining) and they are quite complementary. SAXS can provide full 3D orientation measurements on intact tissue, but it operates at a mesoscopic resolution and requires access to a synchrotron. SLI can measure the orientations of multiple fascicles per voxel at a microscopic resolution and relies on more widely accessible equipment, but its accuracy suffers for fiber orientations perpendicular to the imaging plane and it requires …
Reviewer #2 (Public Review):
This work is a cross-validation of an x-ray tomography technique (SAXS) and an optical microscopy technique (SLI) for imaging axonal orientations ex vivo. These innovative methods were introduced in recent papers by the authors, who have teamed up here to compare them side-by-side on the same tissue samples for the first time. The two methods are both label-free (do not require staining) and they are quite complementary. SAXS can provide full 3D orientation measurements on intact tissue, but it operates at a mesoscopic resolution and requires access to a synchrotron. SLI can measure the orientations of multiple fascicles per voxel at a microscopic resolution and relies on more widely accessible equipment, but its accuracy suffers for fiber orientations perpendicular to the imaging plane and it requires tissue to be sectioned before it is imaged. Therefore it makes a lot of sense to explore the complementary strengths of these two techniques, and to use one to "fill in the blanks" of the other. The paper also compares the orientation measurements obtained with SAXS and SLI to those obtained with diffusion MRI. The latter provides only indirect measurements based on water diffusion, at a mesoscopic resolution somewhat lower than that of SAXS, but has the benefit of being feasible in vivo.
A limitation of this study is that conclusions on the comparison between SAXS and SLI are drawn from only 2 sections of a partial monkey brain sample and 2 sections of a partial human brain sample. Conclusions on diffusion MRI are drawn only on the 2 human sample sections. This is particularly an issue for the comparison to diffusion MRI, as the diffusion MRI voxels are wider than the section thickness, hence one cannot preclude that any orientations detected with diffusion MRI but not with SAXS and SLI come from the portion of the voxel that is missing from the corresponding SAXS/SLI section.
The stated aim of the paper is to provide a framework for combining the complementary benefits of SAXS and SLI, rather than simply presenting the results of a cross-validation study. This is a significant and ambitious aim. However, in order for this to serve as a framework, there would have to be clear prescriptions for how researchers interested in obtaining ground-truth measurements of axonal orientations would do so by using these two methods in tandem. This is not adequately developed in the paper in its present form. For example, the results show reasonable agreement between SAXS and SLI orientations when fibers lie within the SLI imaging plane and decreasing agreement for fibers with increasing through-plane inclination. How would the two methods be combined in voxels where they disagree? Would one use SLI orientations in voxels with fewer through-plane fibers and SAXS orientations in voxels with more through-plane fibers? How would voxels be assigned to each category? How would the orientation vectors from the two modalities be composed and how would the resolution difference between the two be handled? When the through-plane measurement of SLI is unreliable, is its in-plane measurement still reliable? That is if there were one mainly in-plane and one mainly through-plane fiber population, would the orientation of the former still be measured correctly by SLI? There is also considerable agreement reported here between through-plane orientations obtained with SAXS and diffusion MRI. Would this mean that diffusion MRI itself could be used to supplement SLI with through-plane orientations? Any clear set of prescriptions along these lines would represent a framework for imaging orientations by combining modalities. This, however, would require detailed steps for how to perform the combination and use the multi- vs. uni-modal framework to reconstruct connectional anatomy.
A key advantage of SAXS is that it can be performed on intact samples, i.e., before any nonlinear distortions of the tissue are introduced by sectioning. Thus it can provide an undistorted reference, with contrast on axonal orientations that would be absent in, say, a structural MRI of comparable resolution. This contrast could be used to drive registration of the distorted SLI sections to an undistorted SAXS volume, and therefore is a key way in which the two techniques can complement each other. Here, however, this is not explored, as SAXS is performed after sectioning. It is not clear if this is the authors' prescription for how a combined SAXS/SLI framework would be implemented, or if it was done specifically for this study. First, it would seem that SAXS on the intact sample would be lower maintenance, requiring less setup time and hence potentially less overall beamtime than performing SAXS on each section separately. This would make it more practical for routine deployment beyond a few sections. Second, because the SAXS data are now nonlinearly distorted, they cannot be affinely aligned to the MRI volumes. While, in principle, performing both SAXS and SLI on the sections may facilitate the comparison between the two, having to unmount, rehydrate, and remount the sections in between may negate this advantage, as now there is no guarantee that SAXS and SLI can be affinely registered to each other. Here all these registration steps are performed affinely, so it is unclear to which extent the computed errors between modalities are characterizing the inherent limitations of the respective contrasts, or limitations of the registration technique. Some of the alignment is performed manually, for example, specific regions of the images are realigned by hand, and the slice of the diffusion MRI volume that is aligned to the SAXS/SLI sections is chosen by hand. Again, for this to serve as a framework that can be deployed on whole samples, there would have to be clear prescriptions for how to perform these steps robustly, how to ensure that the MRI can be acquired in a coordinate frame parallel to the sections, etc.
Finally, the paper puts forth a general conclusion that diffusion MRI overestimates the number of fiber populations per voxel, on the basis of small ODF peaks appearing perpendicular to the main ODF peaks. Of all conclusions in the paper, this is the least convincingly supported by evidence. First, these small perpendicular peaks are a known artifact, which would be typically eliminated by ignoring ODF peaks below a certain amplitude, a common practice in diffusion tractography algorithms. The authors refrain from using an amplitude threshold, with the rationale that it may also remove true diffusion orientations. However, they apply a threshold when they detect SLI peaks (a rather stringent 8% of the maximum). Second, the explanation that these artifactual peaks may appear due to vessel walls is not convincing. Vasculature is sparse. A single vessel wall will not impact the diffusion signal in the same way as a bundle of parallel axons. In an axon bundle, water molecule displacements are restricted in all directions except parallel to the axons. A single vessel wall in a voxel will not have the same effect on displacements (which are much smaller than the size of the voxel). From Figure 5, it looks like there would be at most 1-2 of these vessels in a diffusion MRI voxel, and they would not be in all voxels. This cannot explain the widespread appearance of these small artifactual peaks. Third, many ODF reconstruction methods have parameters that can be adjusted to make these artifactual peaks more or less prominent. The default parameters may be optimal for in vivo but not ex vivo data, due to the effects of fixation. In light of these concerns, I would caution against making such a general statement about all diffusion MRI in the human brain, especially on the basis of a single diffusion reconstruction method applied to a single location in one brain.
-