Subjects
Thirty-two canine subjects were prospectively recruited from client populations at the Colorado State University Veterinary Teaching Hospital, Fort Collins, CO, USA, between 8/19/2021 and 2/10/2023. All aspects of this study were approved by the Colorado State University Institutional Animal Care and Use Committee and the Clinical Review Board (IACUC protocol numbers: 4106 and 3384). There were 16 castrated males and 16 spayed females from 17 different breeds: American Pit Bull Terrier (n=1), Australian Shephard (2), Blue Tick Hound (1), Border Collie (2), Boxer (1), Chesapeake Bay Retriever (1), Cocker Spaniel (2), Dachshund (1), German Shepherd (2), German Shorthaired Pointer (1), Labrador Retriever (2), Miniature Poodle (1), Pug (3), Rhodesian Ridgeback (1), Shih Tzu (1), Weimaraner (1), and mixed breed (9). Due to the limited nature of literature regarding diffusion tensor imaging of CCD, priori sample size calculation could not be performed. Results of this study can guide future sample size calculations.
The participants were divided into two groups based on neurological examination and questionnaire scoring. A previously validated owner-directed survey, the Canine Dementia Scale (CADES), was used to determine the cognitive health of the dogs. The CADES questionnaire contains 17 questions related to changes in dogs’ behaviour, including spatial orientation, social orientation, social interactions, sleep-wake cycles, and house soiling (23). Normal aging scores range between 0–7 points, mild cognitive impairment between 8–23 points, moderate cognitive impairment between 24–44 points and severe cognitive impairment between 45–95 points (24). Dogs with scores >8 on the CADES surveys were categorized as dogs as having some signs of CCD, while dogs with scores between 0-7 were categorized as cognitively healthy. All dogs were greater than 8 years of age. Neurological examination and anatomical magnetic resonance imaging (MRI) analysis to rule out cognitive decline due to structural pathology were performed by a board-certified veterinary neurologist (S.M) or a veterinary neurology resident (M.U.), both of whom were blinded to the CADES scores. There were no potential confounders.
Magnetic Resonance Imaging (MRI) Acquisition
Magnetic Resonance Imaging of the dogs was performed under general anaesthesia by a board-certified veterinary anesthesiologist. Animals were premedicated with methadone (0.1-0.5 mg/kg) or butorphanol (0.1-0.5 mg/kg). Preoxygenation and monitoring equipment was placed prior to induction (as tolerated). Propofol (2-5 mg/kg) and midazolam (0.1-0.2 mg/kg) were administered intravenously. Mechanical ventilation began immediately after induction to prevent an increase in carbon dioxide. They were maintained using isoflurane or sevoflurane with continuous propofol infusion (100-400 mcg/kg/min). Heart rate and rhythm, body temperature, direct arterial blood pressure, end-tidal carbon dioxide and oxygen saturation levels were monitored throughout the procedure. MRI was performed using a 3.0T Siemens Skyra (Siemens Healthcare GmbH, Erlangen, Germany) whole-body scanner (70 cm bore diameter) operating at 45 mT/m amplitude at 200 T/m/s slew-rate. The subjects were placed in dorsal recumbency with their head centered in a 15-channel receive-transmit knee coil. Diffusion tensor images were acquired in the transaxial plane (TR=3120, TE=112.40, flip angle=78°, isometric voxel size of 1.64 ´ 1.64 ´ 1.90 mm, in-plane field of view=213 mm, matrix size 130 mm ´ 130 mm with 138 gradient directions, b=0 and 2000 s/mm2).
Diffusion MRI Preprocessing
Diffusion weighted imaging (DWI) data were corrected for eddy distortion and motion using eddy_correct (25) in FMRIB’s Software Library (FSL) (26). Brain masks were created using the Brain Extraction Tool (BET) (27) in FSL to separate the brain from surrounding tissues, using the indices of -m, -f 0.85 and -R.
A DTI model was fitted to the data using FMBRIB’s Diffusion Toolbox 5.0 in FSL, creating voxelwise tensor maps for fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD) and axial diffusivity (AxD) (28, 29). FA is the measure of directionality of diffusion in a tissue, with values ranging from 0 (isotropic) to 1 (anisotropic). In white matter, anisotropic diffusion is due to the presence of membranes, as well as myelination (30). The MD reflects the magnitude of the total diffusion in a voxel, regardless of directionality. AxD is the diffusion parallel to the primary diffusion direction, while RD is the diffusion perpendicular to the primary diffusion direction (12, 31).
All images were visually inspected at each step of image processing. Diffusion tensor images that had severe distortion were excluded from the study due to difficulty and uncertainty in identifying the targeted brain region of interest. Three samples were excluded due to severe dorsal-ventral distortion of image.
Region of Interest Definition
Prior to region of interest (ROI) selection, the study was blinded and randomized. Clinical Trials Coordinator was aware of group allocation throughout the study and unblinded data post MRI analysis. To overcome the anatomical and brain size variations across the dog subjects, manual ROI selection was performed using the Fsleyes edit mode (32). The following regions were considered for analysis due to histopathological changes observed in dogs with CCD (33): thalamus, hippocampus, corpus callosum, fornix, and cortex. Image distortion and resolution inhibited consistent delineation of the hippocampus, fornix, and cortical regions. Therefore, the splenium of the corpus callosum and thalamus were selected as ROIs because of feasibility of voxel selection based on consistent MRI quality/resolution. Veterinary neurology resident (M.U.) and veterinary neurology research assistant (J.H.) individually selected voxels for the ROIs for each image sample, and after all analyses were completed, the average of the two ROIs for each image was used. Analysis of thalamic regions in two animals were excluded due to distortion of the thalamus inhibiting effective delineation.
Four voxels from the splenium of the CC and eight voxels from the thalamus were included in the ROI to minimize contamination from extraneous structures while maximizing the data points included within the mean of each parameter. Transverse sections of the FA map were used for visualization of structures, and color FA map was used to enhance visibility of axonal tracts and avoid contamination of the internal capsule when selecting voxels of the thalamus ROI and ventricles when selecting voxels for the CC ROI (34). Figure 1 shows an ROI of the CC and Figure 2 an ROI of the thalamus using black and white and color FA maps. The mean data were extracted from the ROIs using fslmeants. The mean FA, MD, AxD, and RD were acquired for the CC and thalamus of each subject.
Statistics
The statistical analyses were performed by statistician (T. Z.) and veterinary neurology resident (M.U.). Interobserver analysis of the ROIs was performed by calculating the intraclass correlation coefficients between the observers and their 95% confidence intervals, shown in Table 2. Values less than 0.5 are indicative of poor reliability, values between 0.5 and 0.75 indicate moderate reliability, values between 0.75 and 0.9 indicate good reliability, and values greater than 0.90 indicate excellent reliability.
Table 2: Intra-observer analysis comparing subjective data collection from two regions.
|
Corpus Callosum
|
Thalamus
|
DTI Index
|
ICC
|
ICC.Lower
|
ICC.Upper
|
ICC
|
ICC.Lower
|
ICC.Upper
|
FA
|
0.851
|
0.720
|
0.924
|
0.387
|
0.041
|
0.651
|
MD
|
0.890
|
0.788
|
0.944
|
0.636
|
0.367
|
0.808
|
AxD
|
0.831
|
0.685
|
0.914
|
0.508
|
0.191
|
0.730
|
RD
|
0.878
|
0.767
|
0.938
|
0.587
|
0.298
|
0.779
|
Abbreviations: FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; ICC, intraclass correlation coefficient
All collected DTI variables (FA, MD, AxD and RD) were evaluated for normality. The Wilcoxon rank sum test was performed to evaluate whether there was difference in the median of each DTI variable between the CCD and healthy groups.
Spearman’s rank correlation rho was performed to analyse the correlation between the CADES score and the mean FA, MD, AxD and RD values. All the statistical analyses were conducted using R statistical software (R Core Team, 2023).
Post hoc power analysis based on differences in the DTI parameters between the CCD and healthy groups was performed separately for the CC and thalamus and the four DTI parameters (FA, MD, AxD and RD). The significance level for the comparisons was 0.05. Results of post hoc power analysis can be seen in Table 3.
Table 3: Post hoc power analysis for DTI Indexes between CCD and healthy groups
|
Corpus Callosum
|
Thalamus
|
DTI Index
|
Power
|
n.80
|
n.90
|
Power
|
n.80
|
n.90
|
FA
|
0.1423
|
144
|
192
|
0.0690
|
598
|
800
|
MD
|
0.2589
|
66
|
87
|
0.5566
|
24
|
31
|
AxD
|
0.0709
|
613
|
821
|
0.2742
|
54
|
72
|
RD
|
0.2154
|
62
|
110
|
0.4879
|
28
|
37
|
Abbreviations: DTI, diffusion tensor imaging; CCD, canine cognitive dysfunction; FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity