In vivo mapping of striatal neurodegeneration in Huntington’s disease with Soma and Neurite Density Imaging

Reviewer #1 (Public review):

(1) In this study, the authors aimed at characterizing Huntington's Disease (HD) - related microstructural abnormalities in the basal ganglia and thalami as revealed using Soma and Neurite Density Imaging (SANDI) indices (apparent soma density, apparent soma size, extracellular water signal fraction, extracellular diffusivity, apparent neurite density, fractional anisotropy and mean diffusivity).

(2) The study implements a novel biophysical diffusion model that extends up-to-date methodologies and presents a significant potential for quantifying neurodegenerative processes of the grey matter of the human brain in vivo. The authors comment on the usefulness of this technique in other pathologies, but they exemplify it only with multiple sclerosis. Further development of this, building evidence, should be provided.

(3) The study found that HD-related neurodegeneration in the striatum accounted significantly for striatal atrophy and correlated with motor impairments. HD was associated with reduced soma density, increased apparent soma size, and extracellular signal fraction in the basal ganglia, but not in the thalami. Additionally, these effects were larger at the manifest stage.

(4) The results of this work demonstrate the impact of HD on the basal ganglia and thalami, which can be further explored as a non-invasive biomarker of disease progression. Additionally, the study shows that SANDI can be used to explore grey matter microstructure in a variety of neurological conditions.

Reviewer #2 (Public review):

Summary:

The authors aimed to investigate whether advanced microstructural diffusion MRI modeling using the SANDI framework could reveal clinically relevant tissue alterations in the subcortical structures of individuals with Huntington's disease (HD). Specifically, they sought to determine if SANDI-derived parameters-such as soma density, soma size, and extracellular diffusivity-could detect abnormalities in both manifest and premanifest HD stages, complement standard MRI biomarkers (e.g., volume, MD), and correlate with disease burden and motor impairment. Through this, they hoped to demonstrate the feasibility and added biological specificity of SANDI for early detection and characterization of HD pathology.

Strengths:

(1) Novelty and relevance:

This is, to the best of my knowledge, the first clinical deployment of SANDI in HD, offering more biophysically interpretable and specific imaging biomarkers than standard DTI or volumetric features.

(2) More specific microstructural insight: Traditional approaches have used volumetric features (e.g., striatal volume loss) or DTI metrics (like FA and MD), which are indirect and non-specific markers. They can indicate something is "wrong" but not what is wrong.

(3) SANDI parameters permit establishing clearer links with microstructure:

o Apparent soma density (fis): proxy for neuronal/glial cell body density.

o Apparent soma size (rs): reflects possible gliagl hypertrophy or neuronal shrinkage.

o Neurite density (fin): linked to dendritic/axonal integrity.

o Extracellular fraction and diffusivity: sensitive to edema, gliosis, and tissue loss.

In this way, a decrease in soma density can be related to neural loss (e.g., medium spiny neurons), and an increase in soma size and extracellular fraction could be related to glial reactivity (astrocytes, microglia). This enables differentiating between atrophy due to neuron loss vs reactive gliosis, which volumetrics or DTI cannot do.

(4) Integration of modalities: The inclusion of motor impairment (Q-Motor), HD-ISS staging, and multi-compartment diffusion modeling is a methodological strength.

(5) Early detection potential: SANDI metrics showed abnormalities in premanifest HD, sometimes even when volume loss was mild or absent. This suggests the potential for earlier, more sensitive biomarkers of disease progression.

(6) Predictive power: Regression models showed that SANDI metrics explained up to 63% of the variance in striatal volumes in HD. And this correlated strongly with motor impairment and disease burden (CAP100). This shows they are not just redundant with volume or DTI, but they are complementary and potentially more mechanistically meaningful.

Weaknesses:

Certain aspects of the study would benefit from clarification:

(1) Scanner and acquisition consistency: While HD data are from the WAND study, it is not clear whether controls were scanned on the same scanner or protocol. Given the use of model-derived metrics (especially SANDI), differences in scanner or acquisition could introduce confounds. Also, although it offers novel and biologically informative markers, widespread clinical translation still faces hurdles. For instance, the study used a 3T Connectom scanner (300mT/m gradients), which is not widely available. Reproduction of these results in standard 3T clinical scanners would be a great addition, in scenarios with lower resolution, less precise parameter recovery, and longer scans if SNR needs to be maintained.

(2) HD-ISS staging and group comparisons:
a) Only 26-27 out of 56 gene-positive participants could be assigned HD-ISS stages, and none were classified into stages 0 or 4.

b) Visual overlap between stages 1 and 2 in behavioral and imaging features suggests that staging-based group separation may not be robust.

c) The above may lead to claims based on progression across HD-ISS stages to be overinterpreted or underpowered

(3) Regression modeling choices:
a) SANDI metrics included in the models differ between HC and HD groups, reducing comparability.

b) The potential impact of multicollinearity (e.g., between fis and rs) is not discussed.

c) Beta coefficients could reflect model instability or parameter degeneracy rather than true biological effects.

These issues do not undermine the study's main conclusions, which effectively demonstrate the feasibility and initial clinical relevance of applying SANDI to HD. Nonetheless, addressing them more thoroughly would enhance the clarity and interpretability of the manuscript.

Reviewer #3 (Public review):

Summary:

Ioakeimidis and colleagues studied microstructural abnormalities in N=56 Huntington's disease (HD) patients compared to N=57 normative controls. The authors used a powerful MRI Connectom scanner and applied the SANDI model to estimate the soma size, neurite size, soma density, and extracellular fraction in key subcortical nuclei related to HD. In the striatum, they found decreased soma density and increased soma size, which also seemed to become more pronounced in advanced HD individuals in the final exploratory analyses. The authors conducted important analyses to find whether the SANDI measures correlate with clinical scores (i.e., QMotor) and whether the variance of the striatal volume is explained by the SANDI measures. They found a relationship between SANDI measures for both.

Strengths:

The study is both innovative and of high interest for the HD community. The authors provide a rich pool of statistical analyses and results that anticipate the questions that may emerge in the HD research community. Statistics are carefully chosen and image processing is done with state-of-the-art methods and tools. The sample size gives sufficient credibility to the findings. Altogether, I think this study sets a milestone in the attempts of the HD community to understand neuropathological processes with non-invasive methods, and extends the current knowledge of microstructural anomalies identified in HD with diffusion MRI. More importantly, the newly identified anomalies in soma size and soma density open new avenues for studying these biological effects further and perhaps developing these biomarkers for use in clinical trials.

Weaknesses:

(1) An important question is whether the SANDI measures, which require an expensive scanner and elaborate processing, are better biomarkers than the more traditional DTI measures. Can the authors compare the effect size of FA/MD with SANDI measures? In some of the plots and tables, FA/MD seem to have comparable, if not higher, correlations with QMotor or CAP scores. On the same vein, it is unclear whether DTI measures were included in hierarchical stepwise regression. I wonder if the stepwise models may have picked up FA/MD instead of SANDI measures if they are given a chance. Overall, I hope the authors can discuss their findings also in this light of cost vs. benefit of adopting SANDI in future studies, which is an important topic for clinical trials.

(2) Similar to the above point, it is very important to consider how strong the biomarking signal is from SANDI measures compared to the good old striatal volume. Some plots seem to indicate that volumes still have the highest correlation with QMotor and the highest effect size in group comparisons. It would be helpful for the community to know where the new SANDI measures stand compared to the most typically used volumes in terms of effect size.

(3) The diffusion measures are inevitably correlated to some degree. Please provide a correlation matrix in the supplementary material, including all DWI measures, to enable readers to better understand how similar SANDI measures are to each other or vs. other DTI measures. Perhaps adding volumes to this correlation matrix may also be a good future reference.

(4) ISS stages:

a) The online ISS calculator requires cut-offs derived from the longitudinal Freesurfer pipeline, while the authors do not have longitudinal data. Thus, the ISS classification might be inaccurate to some degree if the authors used the FS cross-sectional pipeline. Please review this issue and see if updated cut-offs should be used to classify participants.

b) Were there really no participants with ISS 0 among the 56 HD individuals? Please clarify in the manuscript.

(5) A note on terminology that might be confusing to some readers. According to the creators of ISS, the ISS stages are created for research only; they are not used or applied in the clinic. On the other hand, the terms "premanifest" and "manifest" have a clinical meaning, typically based on the diagnostic confidence level. The assignment of ISS0-1 to premanifest and ISS2-3 to manifest may create some non-trivial confusion, if not opposition, in some segments of the HD community. The authors can keep their current terminology, but will need to at least clarify to the reader that this assignment is speculative, does not fully match the clinically-based categories, and should not be confused with similarly named groups in the previous literature.

Author response:

Response to Reviewer 1:

Ad (2) Clinical applications of SANDI have primarily focused on Multiple Sclerosis. However, since the preparation of the manuscript, one study has been published reporting reductions in apparent soma density and white and grey matter differences in apparent soma size in amyotrophic lateral sclerosis (ALS) (https://doi.org/10.1016/j.ejrad.2025.111981). We will include this paper in our revised manuscript.

Responses to Reviewer 2:

Strength:

Ad (3) SANDI cannot directly differentiate between neural and glia cells but the pattern of differences in the SANDI parameters we observed in Huntington’s disease (HD) are consistent with the known pathology in HD.

Weaknesses:

Ad (1) With regards to the question about scanner and acquisition consistency, we can confirm that all diffusion data of individuals with HD and healthy controls from the WAND study were acquired with the same multi-shell High Angular Resolution Diffusion Imaging (HARDI) protocol on the 3T Connectom scanner at CUBRIC. Thus, all diffusion data analysed and reported in this manuscript were acquired with the same protocol on the same strong gradient MRI system for harmonization and consistency purposes.

We agree that for clinical adoption it is important to demonstrate that HD-related SANDI differences do not require ultra-strong gradient imaging and can be detected on standard clinical MRI systems. While we have not collected such data in people with HD, we and others have demonstrated the feasibility of modelling SANDI metrics from multi-shell diffusion-weighted imaging data acquired with maximum b-value 3,000 s/mm2 on clinical 3T MRI system in typical adults and people with MS or ALS (https://doi.org/10.1002/hbm.26416, https://doi.org/10.1038/s41598-024-60497-6, https://doi.org/10.1016/j.ejrad.2025.111981). These studies have demonstrated that it is feasible to characterise brain microstructural differences with SANDI on clinical scanners and that comparable patterns of results can be observed across different MRI systems. It should also be noted that there is presently a move towards stronger gradient implementation in clinical systems as demonstrated by the release of the Siemens Cima.X system which will allow higher b-value diffusion scanning on clinical systems.

ad (2) We agree that due to the small number of HD participants with HD-ISS staging the exploratory comparisons between ISS stages need to be interpreted with caution. We hope to gain access to some of the missing ISS information and plan to include these in the revised paper.

Ad (3) With regards to the queries about the regression modelling choices:

(1) As SANDI metrics differed between HC and HD groups, and hence may not be directly comparable, separate regression models for HC and HD data were conducted without formal comparisons between slopes. Only descriptive exploratory comparisons of the observed pattern were included.

(2) We will provide cross-correlational analyses between all SANDI parameters in the supplements of the revised version of the paper to check for multicollinearity.

(3)All model-based approaches, including SANDI, may be prone to model instability or parameter degeneracy and we will acknowledge and discuss this in the revised version.

Responses to Reviewer 3:

Weaknesses:

Ad (1) and (2) The effect sizes (ES) of group differences in SANDI, DTI, and volume measures in the caudate and putamen (Tables 3 and 4) were broadly comparable: apparent soma radius rs (rrb = 0.45 -0.53), apparent soma size fis (rrb = 0.32 -0.45), FA (rrb = 0.38 -0.55), MD (rrb = 0.51 -0.61) and volumes (rrb = 0.49 -0.55 ). Similar ES were observed between fis and FA, and between rs and volumes. MD showed the largest ES, likely due to striatal atrophy-related CSF partial volume contamination.Cost-benefit analyses of imaging marker choices in clinical trials depend on the aim of the study. DTI provides sensitive but unspecific indices that are influenced by biological and geometrical tissue properties and capture a multitude of microstructural properties. Similarly, volumetric measurements do not inform about the underpinning neurodegenerative processes.

With the advancement of disease-modifying therapies for HD it has become important to identify non-invasive imaging markers that can inform about the mechanistic effects of novel therapies. While DTI and volume metrics are sensitive to detect brain changes, they do not provide specific information about the underpinning tissue properties. Such information, however, may turn out to be important for the evaluation of mechanistic effects of novel therapeutics in clinical trials. Advanced microstructural models such as SANDI may help provide such information. We found that SANDI indices had statistically similar power to the gold standard measures of volumes, but with the added value of information underpinning microstructure. We and others have also shown that SANDI can be applied to multi-shell diffusion data acquired in a clinically feasible time (~10 min) on standard 3T MRI systems (please refer to our response above).

To summarise, DTI and volumes are sensitive to brain changes but will need to be complemented by more advanced microstructural measurements such as SANDI to gain a better understanding of the underlying tissue changes and effects of disease-modifying therapies.

Ad (3) We will provide a correlation matrix of all DWI measures in supplementary material to allow a better understanding how similar SANDI measures are to each other and compared to DTI measures.

Ad (4) Most of the people with HD who have taken part in our study were participants in the Enroll-HD study. We will use HD-ISS information from ENROLL as much as possible. As we do not have longitudinal imaging data for all individuals classified as ISS <2, we will compare our cross-sectional striatal volumes with those from age and sex matched individuals from WAND to determine whether people fall into ISS 0 or 1 category. This approach will hopefully allow us to increase the total HD-ISS sample size and to determine whether there were participants with ISS 0 in our sample.

Ad (5) We will explain in the revised manuscript that ISS stages are created for research only purposes and are not used or applied in clinic, while “premanifest” and “manifest” are helpful concepts in the clinical context. We will clarify that we refer to individuals without motor symptoms as assessed with Total Motor Score (TMS) as premanifest and to those with motor symptoms as manifest. This roughly corresponds to individuals at ISS 0/1 without signs of motor symptoms compared to individuals at ISS 2-3 with signs of motor symptoms.