Change in the Cortical Complexity of Spinocerebellar Ataxia Type 3 Appears Earlier than Clinical Symptoms

Patients with spinocerebellar ataxia type 3 (SCA3) have exhibited cerebral cortical involvement and various mental deficits in previous studies. Clinically, conventional measurements, such as the Mini-Mental State Examination (MMSE) and electroencephalography (EEG), are insensitive to cerebral cortical involvement and mental deficits associated with SCA3, particularly at the early stage of the disease. We applied a three-dimensional fractal dimension (3D-FD) method, which can be used to quantify the shape complexity of cortical folding, in assessing cortical degeneration. We evaluated 48 genetically confirmed SCA3 patients by employing clinical scales and magnetic resonance imaging and using 50 healthy participants as a control group. According to the Scale for the Assessment and Rating of Ataxia (SARA), the SCA3 patients were diagnosed with cortical dysfunction in the cerebellar cortex; however, no significant difference in the cerebral cortex was observed according to the patients’ MMSE ratings. Using the 3D-FD method, we determined that cortical involvement was more extensive than involvement of traditional olivopontocerebellar regions and the corticocerebellar system. Moreover, the significant correlation between decreased 3D-FD values and disease duration may indicate atrophy of the cerebellar cortex and cerebral cortex in SCA3 patients. The change of the cerebral complexity in the SCA3 patients can be detected throughout the disease duration, especially it becomes substantial at the late stage of the disease. Furthermore, we determined that atrophy of the cerebral cortex may occur earlier than changes in MMSE scores and EEG signals.


Introduction
Materials and Methods

Participants
The Institutional Review Board of Taipei Veterans General Hospital approved this study. All participants provided written informed consent before participating. Forty-eight SCA3 patients were recruited from 2005 to 2012. A group of 50 healthy volunteers were matched according to age, sex, and years of education, serving as a control group. An ataxial clinical assessment of all of the patients was conducted using SARA scores and modified Hoehn and Yahr staging [43]. In addition, 17 patients received routine electroencephalography (EEG) examinations and 14 patients were assessed using the MMSE [44]. Table 1 shows the demographic, clinical, and MRI data for both groups. SCA3 patients who experienced progressive and otherwise unexplained ataxia and tested positive for the SCA type 3 genotype were included in this study. The disease duration of the SCA3 patients was 8.89 ± 6.432 years. To further investigate the difference in cortical change according to disease duration, we separated the SCA3 patients into two groups: an early stage group (23 patients of whom the disease duration was shorter than 8 years) and a late stage group (25 patients of whom the disease duration was longer than 8 years).
The members of the control group had no central nervous system disease and exhibited no neurological abnormalities during the study period. An experienced neuroradiologist examined the T1-and T2-weighted images of the control group to ensure the absence of latent neurological diseases or unexpected abnormalities.

Electroencephalography
2.1 Recording. The EEG data were collected when the participants were sitting relaxed and with their eyes closed. Alertness was continually monitored to prevent drowsiness. The EEGs were recorded using a 19-channel digital portable EEG machine (Nicolet Biomedical Inc., Madison, WI) for 5 minutes. All electrodes were placed based on the configuration of the international 10-20 system with a linked ear reference (mean EEG signal of electrodes at bilateral ears). Electrode skin impedance was usually less than 3 kO. The signals were digitized at a rate of 256 Hz by using a 12-bit analogue-to-digital converter and processed using an analogous band-pass filter (0.05−70 Hz) and a 60-Hz notch filter. 2.2 Criteria of the abnormality in EEG. Abnormal EEG patterns of the SCA3 patients were identified by experienced neurologists, which included spikes, intermittent slow activity, diffuse background slowing and frontal and occipital intermittent rhythmic delta activity. Experienced neurologists examined all the EEG recordings without prior informing that these recordings collected from SCA3 patients or other participants.
3.2 Data processing. The acquired T1-weighted image of each participant was reformatted into an axial image and converted to the Analyze format by using MRIcro software (Chris Rorden, University of Nottingham, UK; www.sph.sc.edu/comd/rorden/mricro.html). Fig 1 shows the data-processing flowchart. To improve the accuracy of brain tissue extraction, an automatic skull-stripping function was applied to the image volumes by using the brain extraction tool in the MRIcro software (Fig 1a). Fig 1b shows the normalization process, which was conducted using DiffeoMap (Li, X.; Jiang, H.; and Mori, S.; Johns Hopkins University, www.MriStudio.org). In this procedure, a 12-parameter affine transformation (AIR; Woods, Grafton [45]) was used to normalize each T1-weighted image toward the JHU_MNI_SS_T_ss T1 template.
3.3 Atlas extraction. Image data processing was conducted using the SPM5 toolbox (Welcome Department of Cognitive Neurology, Institute of Neurology, University College London, London, UK, http://www.fil.ion.ucl.ac.uk/spm/) and the Individual Brain Atlases using Statistical Parametric Mapping (IBASPM; http://www.thomaskoenig.ch/Lester/ ibaspm.htm) toolbox in MATLAB 7.0 software (Mathworks, Natick, MA, USA). This process consisted of the following steps: (1) the normalized image volume was segmented into gray matter, white matter, and cerebral spinal fluid in native space (Fig 1c), and (2) the gray matter image was transformed into Montreal Neurological Institute space [46], and each voxel of gray matter was anatomically assigned to one of the 116 automatic anatomical label structures by using IBASPM [46] (Fig 1d). The 26 regions of the cerebellum were then merged into 7 regions according to their anatomical structures, and the volumes of 97 labeled (45 for each cerebral hemisphere; Table 2) brain structures were extracted for each participant (Fig 1e).
3.4 Three-dimensional box-counting method for fractal analysis. Mandelbrot originally proposed fractal analysis to quantify the shape complexity of objects into a single numerical value [47]. Because fractal analysis involves topologically measuring complexity, a high FD value represents a high degree of topological complexity in the examined tissue [48]. In this study, we adopted the 3D box-counting method because it can be used to evaluate the FD of structures with or without self-similarity [26,48]. Thus, we obtained the 3D-FD values of 97 regions of gray matter from the entire brain for each participant. The algorithm of box-counting method was illustrated in our previous studies [22].

Statistical analysis
The normality of each data group was verified using the Jarque-Bera test [49]. The sex and age differences between the groups were measured using the Pearson chi-square test (χ 2 = 0.384, p = 0.535) and the 2-tailed 2-sample t test (p = 0.516), respectively.
We used Pearson's r measurement to investigate the correlations between disease duration and SARA scores and between disease duration and MMSE scores. The 2-tailed t test was used to determine whether the 3D-FD values of each cortical region differed significantly between the control group and the SCA3 group. The significance of the results was corrected according to the false discovery rate (threshold of 0.05) [50]. The magnitude of the association between the 3D-FD value of individual brain regions and clinical features, such as disease duration, SARA scores, and MMSE scores, was determined using Pearson's r measurement. We conducted multiple linear regressions to verify the association between disease duration and 3D-FD values (as a dependent variable), with age controlled. In addition, we explored the differences in the progression rate in each parcellated region of the cerebral cortex by using an ANOVA (with a Bonferroni test for post hoc analyses). These analyses were conducted using the Statistics Toolbox in MATLAB 7.0.

Results
SCA3 patients exhibited cortical dysfunction in the cerebellar cortex (SARA), but no significant difference in the cerebral cortex (MMSE) In addition, 5 of 17 SCA3 patients exhibited abnormal EEG signals in the cerebral hemisphere. Specifically, 3 of 5 SCA3 patients with abnormal EEG recording show intermittent slowing, and 2 of 5 SCA3 patients with abnormal EEG recording exhibited diffuse background slowing. We accordingly classified the SCA3 patients into 2 groups: SCA3 patients with abnormal EEG signals (5 cases) and SCA3 patients with normal EEG signals (12 cases). The disease durations of SCA3 patients with abnormal EEG signals has the trend to be longer than that of SCA3 patients with normal EEG signals (duration for SCA3 patients with abnormal EEG signals: 13.3 ± 5.57; duration for SCA3 patients with normal EEG signals: 6.71 ± 3.51).

Significant correlation between decreased 3D-FD values and disease duration in SCA3
The 3D-FD values of the cerebellar cortex and cerebral cortex of the SCA3 patients exhibited significant decreases compared with those of the control participants (Fig 3a and Fig 4a). We observed a significant correlation between decreased 3D-FD values of both the cerebral and cerebellar cortices and disease duration in the SCA3 patients (Fig 3b and Fig 4b).
Change of the cerebral cortex in SCA3 patients was more substantial at the late stage of the disease The SCA3 patients exhibited a significant correlation between the 3D-FD values of the cerebellar cortex and SARA scores (r = −0.3346; p = 0.023). However, the correlation between the 3D-FD values of the cerebral cortex and MMSE scores in the SCA3 patients was nonsignificant. When we used the 3D-FD method, the SCA3 patients exhibited significantly decreased FD values, even at the early stage. The mean 3D-FD values of the SCA3 patients at the early stage were greater than those at the late stage of the disease (Fig 5a and Fig 5b). This result suggested that cortical atrophy is more apparent at the late stage than at the early stage of the disease.

Widespread involvement of the supratentorial regions in SCA3 patients
3D-FD analysis indicated that 37 parcellated regions of the cerebral and cerebellar cortices exhibited significant regional variation in the SCA3 patients compared with the controls (Table 3). Among the parcellated regions, 32 were located in the cerebral cortex and 5 were located in the cerebellar cortex (Fig 6). Notably, the posterior lobe of the cerebellar cortex show significant decrease in SCA3 patients. In addition, unlike previous studies based on the volumetric method or postmortem analysis [51], our study revealed no significant difference in the 3D-FD values of the bilateral thalamus between the SCA3 patients and the healthy participants. We speculated that the thalamus may shrink but retain its shape, and that the topological complexity of an atrophied thalamus may not change. Consequently, atrophy of the thalamus might not be detected using the FD method.  The progression rates of cortical atrophy in the supratentorial regions were similar In SCA3 patients, we observed a significant correlation (p < 0.05) between disease duration and the 3D-FD values of 37 specific regions, including the anterior and posterior lobes of the cerebellar cortex and frontal lobe, parietal lobe, occipital lobe, temporal lobe, and limbic and subcortical regions of the cerebral cortex. To exclude the effect of age, we conducted multiple regression analysis to verify the association between disease duration (as an independent variable) and 3D-FD values (as a dependent variable). The results indicated that all 37 regions exhibited significant correlation with disease duration when the effect of age was excluded (Fig 7a). In addition, we observed similar progressions of the decreasing 3D-FD value in 32 cerebral regions as the disease duration increased (Fig 7b). No significant difference was observed among these individual regions by using an ANOVA with Bonferroni post-hoc analysis. The degrees of freedom for regions = 98, and F = 0.02.

Cortical change was correlated with cerebellar dysfunction in SCA3
In this study, decreased cortical complexity in cerebellum of SCA3 inferred the cortical atrophies in cerebellum. Similar morphological changes in the cerebellum of the SCA3 patients were reported by previous studies using MRI-based volumetric methods [52,53]. Furthermore, the 3D-FD value in the cerebellar cortex of the SCA3 patients exhibited a significant negative correlation with SARA scores, which demonstrated that the 3D-FD method is effective in quantifying cortical changes and evaluating the association between cortical complexity and dysfunction of the cerebellar cortex in SCA3 patients.

SCA3 patients exhibited a widespread atrophic area in the supratentorial region
Cerebellar cortex, especially the posterior lobe, shows significant decrease in SCA3 patients, and indicated the motor-related impairments in SCA3. Nonetheless, when pathology is in the lateral hemisphere of the posterior cerebellum, the universal cerebellar impairment manifests the cerebellar cognitive affective syndrome (CCAS), including motor-related impairments, executive and visual-spatial impairments, and with affective disturbance [54][55][56]. Previous studies indicated that the clinical features of SCA3 are in accordance with the CCAS [57,58]. Moreover, the involvement of posterior cerebellum was supported by previous pathological studies, which indicated that the middle cerebellar peduncle and central white matter of the cerebellum of SCA3 patients were primarily affected, and the cerebellar cortex and Purkinje were mildly affected [59]. However, the involvement of the posterior lobe of the cerebellar cortex in SCA3 patients has not been observed in previous VBM studies [14,16,60]. Unlike these VBM studies, the results of this study indicated that the cerebellar lobes, including the posterior lobe, are extensively involved in SCA3 patients. These findings are consistent with pathological findings related to SCA3 patients and support the hypothesis that the clinical features in SCA3 were consistent with CCAS [54]. The CCAS results from the disruption of the corticocerebellar system [54,55,61], which is associated with the prefrontal, superior parietal, superior temporal, and limbic cortices [51,62]. Patients with CCAS exhibited impairments in executive functions, shifting, verbal fluency, abstract reasoning, and working memory, all of which are associated with visuospatial deficits, personality changes, and language deficits [51,62]. These cognitive and emotional impairments imply the presence of widespread extracerebellar lesions in the cerebral cortex [63][64][65][66][67]. Our results indicated a significant difference in the supratentorial regions of SCA3 patients, including the frontal, parietal, temporal, occipital, and limbic lobes and subcortical regions. Moreover, a significant negative correlation was observed between the 3D-FD values of these individual regions and disease duration in SCA3 patients. This result revealed that the involvement in SCA3 patients was more widespread than the traditional olivopontocerebellar atrophy pattern [68] and the corticocerebellar system.
FD changes of the cerebral cortex explain why the cerebral-cortexrelated mental decline of SCA3 patients is apparent at the late stage of the disease Our results suggested that cerebral abnormality is more apparent at the late stage of SCA3 because the mean 3D-FD values at the early stage were greater than the late stage. This result is consistent with the presence of dementia at the late stage of SCA3 [69]. Conversely, the MMSE scores show no statistical difference between SCA3 patients and normal controls throughout the disease duration.
Moreover, the observations of abnormal EEG signals in the cerebral hemispheres can provide additional information in supporting cerebral involvement, such as abnormal activities in frontal and temporal regions of schizophrenics [70][71][72] and in parietal and occipital lobe of affective disorders [71,73]. Our results demonstrated abnormal EEG recordings in part of SCA3 patients, namely, intermittent slowing and diffuse background slowing, which were in line with previous SCA3 studies [69,74]. Both the background slowing and slow-wave activity may result from the clinical entities involving cortical regions [75,76]. Besides, EEG signals reflect only superficial cortical activity, these results can be considered direct physiological evidence supporting cortical involvement in SCA3 patients. Moreover, the duration of SCA3 in patients with abnormal EEG signals has the trend to be longer than that in patients with normal EEG signals. This result is consistent with those of previous studies indicating that abnormal EEG signals in SCA3 patients are observed primarily at the late stage of the disease [69,77]. Our results suggested involvement of the cerebral cortex in SCA3 patients throughout the disease duration; however, abnormally neuronal viability of the cerebral cortex of SCA3 patients can be detected at the late stage of the disease using EEG.
In this study, we demonstrated the feasibility of detectible changes of cortical complexity in SCA3 patients. Our results indicate that 3D-FD is the most sensitive measurement for SCA3 in detecting the cerebral cortical involvement and mental associated deficits, especially in early stage of disease duration. Because of limited explanatory power given by the small sample size in EEG recording and MMSE scores of SCA3, we only concluded that (1) the changes of cortical complexity in SCA3 are more substantial than the changes of EEG recording and MMSE scores in SCA3; (2) the cortical involvement in SCA3 patients can be supported by abnormally neuronal viability of the cerebral cortex in SCA3 patients. Besides, our results regarding EEG recording and MMSE scores were in line with previous studies [69,74,78,79].
Another limitation might be the sensitivity of MMSE in detecting the frontal-executive and subcortical dysfunction [80]. However, the MMSE has satisfactory reliability, internal consistency and test-retest reliability. Considering some SCA3 patients were difficult to tolerate a long time examination, MMSE requires only 5 to 10 minutes for setting up and is an easy tool for screening the cognitive impairments. Other specific assessments of cognitive abilities and related mental deficits need further be conducted. It would be helpful to investigate the correlation between the cognitive impairments in SCA3 patients and the changes of cortical complexity of SCA3 patients.
In conclusion, the SCA3 patients in this study exhibited reduced complexity in specific supratentorial regions, which were more extensive than traditional olivopontocerebellar regions and the corticocerebellar system. The cortical atrophy quantified using the 3D-FD method was observed in early-stage SCA3 patients, who exhibited no decreased MMSE scores or abnormal EEG patterns. In addition, cortical complexity changes were more apparent at the late stage of SCA3, indicating that abnormal EEG signals are observed only at the late stage of the disease. Therefore, we suggest that the 3D-FD method is more suitable for detecting cortical atrophy early in SCA3 patients than conventional measurements.

Ethical standards
This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Taipei Veterans General Hospital. All participants provided written informed consent before participating in this study.