Brain diffusion tensor imaging in dogs with degenerative myelopathy

Abstract Background Degenerative myelopathy (DM) in dogs shares similarities with superoxide dismutase 1‐associated human amyotrophic lateral sclerosis (ALS). Brain microstructural lesions are quantified using diffusion tensor imaging (DTI) in ALS patients. Objective Characterize brain neurodegenerative changes in DM‐affected dogs using DTI. Animals Sixteen DM‐affected and 8 control dogs. Methods Prospective observational study. Brain DTI was performed at baseline and every 3 months on DM‐affected dogs and compared to controls. Fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity were calculated on specified regions of interest. Gait scores (0, normal to 14, tetraplegia) were assigned at each scan. Diffusion tensor imaging values in DM‐affected dogs were compared to controls, gait scores, and evaluated over time. Results Mean age was 5.7 years (SD 3.2) in controls and 9.7 years (SD 1.4) in DM‐affected dogs. In DM‐affected dogs, mean baseline gait score was 4 (SD 1), and mean score change from baseline to last scan was 4.82 (SD 2.67). Nine dogs had ≤3 scans; 7 had >3 scans. Accounting for age, no differences in DTI indices were identified for any brain or proximal spinal cord regions between DM‐affected dogs and controls (P > .05). Diffusion tensor imaging values poorly correlated with gait scores (R 2 < .2). No significant changes were identified in diffusion indices over time (P > .05). Conclusions and Clinical Importance Diffusion tensor imaging indices did not differentiate DM‐affected from control dogs, detect longitudinal changes, or differentiate disease severity. Findings do not yet support brain DTI as an imaging biomarker.


| INTRODUCTION
Degenerative myelopathy (DM) is a late adult-onset, progressive neurodegenerative condition in dogs that shares similarities with some forms of superoxide dismutase 1 (SOD1)-associated human amyotrophic lateral sclerosis (ALS). [1][2][3] Degenerative myelopathy causes widespread degeneration of sensory and motor neuronal pathways of both the central and peripheral nervous systems. Definitive diagnosis is attained with histopathology postmortem but presumptive diagnosis is achieved by confirming the SOD1 genotype coupled with compatible clinical signs and advanced imaging to rule out other causes of thoracolumbar myelopathy. 1 Objective disease biomarkers capable of enhancing diagnosis and tracking disease progression are needed to assist in the development of therapeutic interventions for both dogs with DM and people with ALS. Various blood, cerebrospinal fluid, and electrodiagnostic biomarkers are described in DM-affected dogs and people with ALS. [4][5][6][7][8][9] Advanced imaging techniques have also been investigated as potential noninvasive imaging biomarkers in people with ALS with a growing number of explorations in dogs with DM. 4,6,10-12 Diffusion tensor imaging (DTI) is a variation of diffusion-weighted magnetic resonance imaging (MRI) that leverages the cellular motion of water to quantify microstructural lesions, primarily in white matter. 13,14 In the brains of ALS patients, DTI demonstrates widespread white matter changes compared to healthy controls, 4,12,[15][16][17][18][19][20][21][22][23] and longitudinal changes are detectable in some brain areas. 15,16,20,[24][25][26] Corticospinal tract changes are variably correlated with disease progression, validated clinical scales used in ALS patients and electrodiagnostic tests. 16,18,[21][22][23][24][25] Spinal cord DTI in DM-affected dogs identifies alterations in diffusivity parameters in regions with more severe lesion burden compared to controls but is less able to differentiate mildly affected areas. 11 Brain DTI is not reported in DM but there can be histologic changes in the brains and proximal spinal cords of DM-affected dogs.
While 2 case series did not identify overt degeneration in the brain, 27,28 there are variable lesions in the red nucleus, vestibular nuclei, dentate nucleus, nucleus gracilis, and fasciculus gracilis in the cervical spinal cord. 29,30 Brain DTI also detects microstructural changes in other neurodegenerative diseases in dogs. 31,32 This suggests that it is a feasible modality to evaluate brain neurodegenerative changes in dogs with DM. Diffusion tensor imaging of the spinal cord faces several challenges in part because of the small size of the spinal cord contributing to a low signal to noise ratio. 33,34 We speculate that brain DTI might be able to detect lesions in DM-affected dogs complementing what has been reported for spinal cord DTI and provide longitudinal data in these dogs.
The objectives for this project were to use DTI to characterize the location and extent of degenerative changes in the brain of DMaffected dogs compared to healthy controls, and to determine the relationship between DTI indices, clinical disease severity, and disease progression in DM-affected dogs. We hypothesized that DTI would detect degenerative changes in the brain of DM-affected dogs relative to healthy control dogs primarily affecting the white matter regions as well as gray matter areas previously reported to be abnormal histologically in DM-affected dogs. We further hypothesized such changes would correlate with clinical severity and be able to detect disease progression over time.

| Study dogs
Healthy control and DM-affected dogs were prospectively recruited as part of various clinical trials being performed at the University of Missouri Veterinary Health Center. Healthy control dogs were included if they were systemically healthy with no history of neurologic disease, had a neurologic examination that did not reveal abnormalities at the time of imaging, and had a known SOD1 genotype.

| Image processing
Raw diffusion DICOM images were converted to 4D NIfTI format using MRIcron with diffusion direction files (.bvec, .bval) generated by reformatting the PreSet_30axis.tbl file provided by the manufacturer (Canon Medical Systems USA). Using FSL commands, data were corrected for eddy current and motion distortion and an automated mask was used to remove extraneural tissues. A diffusion tensor model was fitted to processed images using the FSL "dtifit" command which provides a matrix-valued tensor for each voxel.
Using T2 weighted transverse images for reference, regions of interest (ROI) were manually created by 1 author for all scans for specific gray and white matter brain regions using the processed images.
We chose to focus on regions where lesions might be present, extrapolated from DTI studies in ALS patients 18,19,21 (eg, body of the corpus callosum, internal capsule) and based on histologic abnormalities that occur in the brains and proximal spinal cord of dogs with DM (eg, red nucleus, dorsal columns). 29,30 Control regions that we anticipated would be unaffected were not specifically targeted. The following areas were initially considered for inclusion: corpus callosum, internal capsule, corona radiata, pre and postcruciate gyrus, caudate nucleus, thalamus, brainstem, red nucleus, caudal cerebellar peduncle, cerebellum (including the deep cerebellar nuclei), proximal spinal cord, dorsal column of proximal spinal cord. Image distortions or artifacts precluded consistently identifying the pre and postcruciate gyrus on all scans and was removed from consideration. Corona radiata, caudate nucleus, thalamus, red nucleus, and brainstem regions were eliminated because of challenges identifying the borders of these regions and consistently applying ROI across dogs and between scans within a given dog. Regions of interest for the body of the corpus callosum, internal capsule, caudal cerebellar peduncle, cerebellum, proximal spinal cord, and dorsal column of the proximal spinal cord were included in the analysis. The size and shape of each ROI was variable but conformed to the structural edges of the specific brain region for the body of the corpus callosum, caudal cerebellar peduncle, and spinal cord (circumferentially and for the dorsal columns). For the internal capsule and cerebellum, the ROI were located within the structure of interest, focusing on creating a region in the same area and of similar size across scans. For all ROI, care was taken to avoid including extraaxial structures when constructing ROI and to be as consistent as possible between dogs and scans. Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) were calculated for each ROI providing quantitative information on the magnitude and direction of diffusivity.

| Statistical analysis
Summary statistics for healthy and DM-affected dogs were reported as mean (SD) or median (range), as appropriate. Normality was determined using the Shapiro-Wilk test. Mean age between the 2 groups was compared using a student's t test. Regression analysis and an ANCOVA incorporating age as a covariate were used to compared mean DTI values for each ROI between DM-affected and control dogs. Among DM-affected dogs, DTI values were compared to gait scores and evaluated over time using Pearson correlation and univariate split-plot approach, respectively. For all analyses, P < .05 was considered significant. The Holm-Bonferroni correction method was used to account for multiple comparisons with P a used to denote adjusted P values.

| DTI in DM-affected compared to control dogs
Fifty-two total scans were performed in DM-affected dogs including 9 dogs with ≤3 scans and 7 dogs with >3 scans (Table 1). Four or more scans were associated with at least 9 months of follow-up after baseline. Mean FA, MD, AD, and RD values for each ROI are listed in Table 2. Accounting for the age difference between groups, significant differences were not identified in any of the DTI indices between DM-affected and control dogs (P > .05, all comparisons).

| DTI and disease stage
Mean duration of neurologic abnormalities (from onset of abnormal signs to initial imaging) was 5.7 months (SD 3.3). All dogs were in disease stage 1 at enrollment. Among the 11 dogs with more than 1 scan, 5 dogs remained in stage 1 throughout, 3 dogs progressed from stage 1 to stage 2, and 3 dogs progressed from stage 1 to stage 3 during the study. Mean baseline gait score was 4 (SD 1), consistent with ambulatory paraparesis. Among the 11 dogs with more than 1 scan, mean change in gait score between baseline and last scan was 4.8 (SD 2.7), representing a progression to nonambulatory paraparesis. Among the 7 dogs with >3 scans, mean baseline gait score was 3.6 (SD 1.5) and mean change from baseline was 6.   This study was not able to identify structural differences in brain and proximal spinal cord regions of DM-affected dogs relative to healthy controls using DTI. While FA in multiple areas was generally lower and MD and RD of the dorsal column were generally higher compared to controls, these differences were not significant. In people with ALS, brain DTI can detect widespread, predominantly white matter degenerative changes compared to healthy controls, even early in the disease course. 4,12,[15][16][17][18][19][20][21][22][23] However, there is variability in which brain areas are reported to be abnormal using DTI, there is notable overlap with values in controls, and postmortem studies confirm unpredictable distribution of lesions in the brain of ALS patients. 12,15,17,36,37 Despite these inconsistencies, DTI is suggested to be potentially useful as a diagnostic biomarker in ALS. 12,15,18 In dogs with DM, there are no other studies of brain DTI and there are relatively few reports detailing brain neurodegeneration. [27][28][29][30] These studies describe either a lack of overt degeneration in the brain 27,28 or lesions variably present in the red nucleus, vestibular nuclei, dentate nucleus, and nucleus gracilis. 29,30 While this suggests that brain histologically well-characterized. 1,27,38 In the only study of spinal cord DTI in DM, FA is significantly lower in severe DM cases compared to control dogs, but only within areas with more severe lesion burden (caudal cervical, mid-thoracic, and cranial lumbar spinal cord). 11 Consistent with our findings, no differences in FA or MD are reported for ROI in the cranial cervical spinal cord in that study. 11 Additionally, FA

| DTI and disease progression
is not different between more mildly affected dogs and healthy control dogs other than in the cranial lumbar spinal cord, and overlap between groups is large. 11 Using brain DTI (including the proximal spinal cord) as a diagnostic biomarker might, therefore, not be useful in dogs with DM due to lack of notable brain neurodegeneration or, if present, limited observed changes in the setting of early-stage disease.
The majority of our healthy control dogs were heterozygous or homozygous affected by the SOD1 mutation. Studies in asymptomatic SOD1 mutation carriers in people report conflicting results, 39,40 but abnormalities in DTI indices consistent with neurodegeneration in the internal capsule are described in people with no clinical signs of ALS. 39 Additionally, in patients diagnosed with ALS, DTI detects microstructural abnormalities in brain regions that remain clinically silent (ie, for which there are no clinical signs of dysfunction). 15  In the current study, FA of the internal capsule decreased over time but the change was no longer significant once corrected for multiple comparisons. We had only 7 dogs that were followed for at least 9 months and marked clinical progression during the study period was uncommon. There was also high interindividual variability and the longitudinal changes in DTI values were small within a given dog. These factors might have impacted our ability to detect changes over time.
Furthermore, there are no imaging-based, longitudinal studies in DM in dogs to which to compare our results. In people with ALS, there is variability between studies regarding associations between DTI and disease progression. In several longitudinal studies, FA decreases, especially in the internal capsule, and more extensive regions of the brain become affected over time. 15,16,20,[24][25][26] However, changes over time are inconsistent when considering group (pooled) versus individual patient data and are typically only significant when considering scans performed many months (>8 months) from baseline. 15,16,26 It is also noted that changes in gray matter regions, in particular cerebellar gray matter, show greater progression over time compared to white matter. 15 These studies suggest that ALS patients must be monitored over a relatively long timeframe for DTI evidence of progression to become apparent and that progression varies between individuals, the type of brain tissue, and the specific brain region. The current study focused primarily on white matter regions, but it is possible progressive neurodegeneration would have been more apparent in gray matter regions as has been reported by people. 15 As noted above, the lack of detectable longitudinal changes might also reflect the differences in lesion burden between ALS patients and DM-affected dogs.
It is possible that brain lesions in dogs are not substantial or progressive enough to be detected via DTI.  ing might prove more relevant in DM. We also used an ROI-based approach to calculate DTI indices, but another analysis method such as tract-based spatial statistics might be more useful, especially when analyzing multiple scans performed over time. It is also possible that combining DTI with other MRI techniques (eg, functional MRI, magnetic resonance spectroscopy, etc.) or functional modalities (eg, transcranial magnetic stimulation) might enhance the sensitivity to identify structural and functional changes DM-affected dogs as has been described in people with ALS. 17,40 Overall, our preliminary results showed that DTI failed to identify overt neurodegenerative changes in the brain and proximal spinal cord of dogs with DM, perhaps due to minimal or absent neurodegeneration in these regions. While brain DTI might have some utility in disease progression or more advanced disease stages, findings do not currently support this modality as a noninvasive, imaging biomarker in dogs with DM.