SPINS CCA analysis
Introduction
Neurocognitive and social cognitive deficits are pervasive in schizophrenia (SSD). Both deficits are often evident early in illness course, highly predictive of functional outcome, and largely unalleviated by existing pharmacological intervention. A great deal of research has described the unique and shared variance of neurocognition and social cognition deficits, as well as their respective, and mutual, relationship to other hallmark features in SSD. More recently, interest in biologically-driven detection and targeted intervention has driven a smaller body of work to investigate the neurobiology underlying neurocognition and social cognition deficits. To date, most imaging studies have tested the association of a single construct (i.e., either neurocognition or social cognition, often considered as unitary rather than dimensional) with whole-brain derived features. As a result, it is unknown if neurocognitive and social cognitive deficits arise from overlapping or distinct brain pathology.
Objective. In the present analysis, we apply a multivariate statistical technique, canonical correlation analysis (CCA), to illuminate the relationship between white matter fractional anisotropy (FA) and neurocognition and social cognition, across a spectrum of SSD and healthy controls (HCs). In past work, we have found FA values in the right AF, left UF, and right ILF to be highly related to social cognition (Voineskos, 2013; Wheeler, 2015; Behdinan, 2015). Here, we seek to replicate this finding, confirm its hemispheric sensitivity, and investigate if these regions are also implicated in neurocognition. We expect to find that white matter disintegrity in these three tracts (and others implicated in social cognition) will also predict neurocognitive impairment, though perhaps only in some facets, and likely to a lesser degree.
Methods
Data were extracted from the collaborative multi-centre study, ‘Social Processes Initiative in Neurobiology of the Schizophrenia(s)’ (SPINS), designed to identify neural circuitry related to social cognitive impairment, in nearly 500 adults who either have SSD or no psychiatric diagnosis (HC). (See Figure 1 for a visualization of available data.) The study has an extensive multi-modal imaging protocol, as well as ‘deep phenotyping’ of neurocognitive and social cognitive ability, from multiple days of participant assessment.
DWI acquisition and preprocessing. The present study analyzed diffusion weighted imaging (DWI) scans, which are sensitive to white matter microstructure. Despite prospective efforts to harmonize DWI acquisitions across sites, analyses revealed that derived metrics showed a large, non-linear scanner effect (i.e., different scanner models produced signals that varied nonlinearly as a function of brain tissue), and we have therefore elected to analyze only data collected from three matched 3T Siemens PRISMA MRIs with 64-channel head-coils. All DWI scans were a 60-direction EPI dual spin echo sequence with the following identical parameters: b=1000, 5 b0s, TR/TE=8800/85ms, FOV=256mm, 128x128 matrix, and 2mm isotropic voxels. DWI data were denoised, skull-stripped, aligned, and motion- and distortion-corrected using FSL and MRtrix software. Data underwent automated and manual quality control after every preprocessing step.
White matter tract estimation. We used the Slicer whitematteranalysis computational pipeline to fit a tensor and perform deterministic whole brain fiber tracking. Streamlines were bundled via registration to the ORG atlas (Zhang 2018), and parcellated into 74 white matter tracts (41 unique; see Supplementary Table 1). For each of these tracts, we estimated fractional anisotropy (FA), which indicates the coherence with which water molecules diffuse along tissue, and is held to be a proxy of white matter integrity (values closer to 1 are indicative of healthy tissue; those closer to 0 indicate probable pathology).
Neurocognitive and social cognitive assessment. Neurocognitive status was assessed via the MATRICS (Measurement and Treatment Research to Improve Cognition in Schizophrenia) Cognitive Consensus Battery (MCCB), which provides an evaluation of key cognitive domains relevant to SSD. In this analysis, we included factor scores for Processing speed, Attention & vigilance, Working memory, Verbal learning, Visual learning, and Problem solving. Social cognitive ability was captured by total scores for the Reading the Mind in the Eyes Test (RMET), Relationships Across Domains (RAD), the Penn Emotion Recognition Test (ER-40), the Interpersonal Reactivity Index (IRI), and the Empathic Accuracy task (EA), and factor scores for the Awareness of Social Inference Test-Revised (TASIT). Importantly, all neurocognitive and social cognitive assessments were administered experimentally, and produce objective performance-based outcomes. Moreover, though all assessments are sensitive to impairment in SSD, a wide distribution of performance is evident across the SSD-HC continuum (no floor or ceiling performance).
Canonical Correlation Analysis. CCA reveals multivariate patterns of linked dimensions between two sets of variables, often denoted as an \(X\) (predictor) set and \(Y\) (criterion) set. Our \(X\) set was comprised of 72 reliably-segmented white matter tracts. The \(Y\) set was comprised of 14 behavioural variables: 6 neurocognition variables, and 8 social cognition variables, described above.
Results
Participants. Participant characteristics are presented in Table 1. Data from 162 participants (78 HC, 84 SSD) were available for analysis after excluding participants for failed eligibility, poor DWI data quality, and excessive missing observations (defined as ≥15/74 FA values (n=17 participants), ≥2/6 neurocognition scores (n=2), or ≥2/8 social cognition scores (n=0)). Missing data from participants with below threshold missingness for the \(X\) (n=1503 tracts) and \(Y\) (n=7 neurocognition, n=27 social cognition) sets were imputed via multivariate chained equations (van Buuren, 2019). Note that the 86 variables across the 162 participants exceeded the ideal 10:1 observation-to-variable ratio guideline for CCA, but not so greatly that dimensionality reduction techniques (e.g., PCA) were thought warranted to avoid over-fitting. In general, our sample was gender-balanced, young adult, well educated, and the SSD participants were clinically stable.
Table 1. Participant characteristics (means and standard deviations).
| Combined, n=162 | HC, n=78 | SSD, n=84 | p | |
|---|---|---|---|---|
| Demographics | ||||
| Sex (F:M) | 63:99 | 39:39 | 24:60 | .008 |
| Age (years) | 32.19 (10.03) | 33.17 (10.37) | 31.29 (9.68) | .234 |
| Handedness (0=L, 1=R) | 0.67 (0.45) | 0.62 (0.50) | 0.72 (0.39) | .176 |
| Education (years) | 15.05 (2.35) | 16.15 (2.08) | 14.02 (2.11) | .000 |
| BMI | 27.60 (5.55) | 26.69 (5.72) | 28.44 (5.29) | .045 |
| WTAR | 109.36 (13.19) | 113.03 (11.13) | 105.95 (14.07) | .001 |
| Neurocognition (MATRICS MCCB) | ||||
| Processing speed | 45.93 (14.50) | 54.27 (9.82) | 38.19 (13.88) | .000 |
| Attention and vigilance | 44.13 (14.00) | 47.96 (14.46) | 40.57 (12.64) | .001 |
| Working memory | 44.37 (11.59) | 46.64 (12.16) | 42.26 (10.68) | .016 |
| Verbal learning | 45.34 (11.35) | 50.99 (10.59) | 40.10 (9.39) | .000 |
| Visual learning | 42.62 (13.25) | 48.13 (11.45) | 37.51 (12.82) | .000 |
| Reasoning and problem solving | 46.42 (11.65) | 49.55 (10.09) | 43.51 (12.29) | .001 |
| Social cognition | ||||
| ER-40 | 3461.12 (546.80) | 3695.38 (368.08) | 3243.59 (595.67) | .000 |
| RMET | 25.68 (5.15) | 27.05 (4.16) | 24.40 (5.65) | .001 |
| EA | 0.47 (0.18) | 0.53 (0.16) | 0.41 (0.19) | .000 |
| RAD | 55.90 (8.59) | 60.36 (5.59) | 51.76 (8.84) | .000 |
| IRI | 66.51 (12.38) | 65.76 (11.69) | 67.21 (13.01) | .455 |
| TASIT 1 | 23.40 (3.17) | 24.78 (2.27) | 22.12 (3.35) | .000 |
| TASIT 2 | 50.75 (7.72) | 54.51 (4.40) | 47.26 (8.50) | .000 |
| TASIT 3 | 50.94 (8.23) | 54.31 (5.16) | 47.81 (9.28) | .000 |
| MRI | ||||
| Absolute motion | 0.49 (0.41) | 0.50 (0.43) | 0.48 (0.40) | .755 |
| Composite image quality score | -0.93 (1.39) | -1.12 (1.43) | -0.75 (1.34) | .095 |
| Whole brain fractional anisotropy (FA) | 0.37 (0.01) | 0.38 (0.01) | 0.37 (0.01) | .002 |
| Clinical, functional, outcome | ||||
| SPQ-B | 5.11 (5.56) | 1.47 (2.52) | 8.49 (5.48) | .000 |
| CIRG-S | 2.81 (3.00) | 1.59 (2.00) | 3.95 (3.31) | .000 |
| BPRS | – | – | 24.88 (11.67) | – |
| SAS | – | – | 556.35 (450.68) | – |
| QLS | – | – | 31.18 (7.82) | – |
| SANS | – | – | 2.21 (2.6) | – |
| Medication | ||||
| CPZE equivalents (mg / day) | – | – | 69.96 (20.23) | – |
| Note: | ||||
|
The BPRS, SAS, QLS, and SANS were only administered to participants with SSD; All neurocognition scores are MATRICS MCCB factor scores and have been normed for age and sex (Nuechterlein et al., 2008); HC = healthy control, SSD = schizophrenia spectrum disorder, BMI = body mass index, WTAR = Wechsler Test of Adult Reading, SPQ-B = Schizotypal Personality Questionnaire-Brief, CIRS-G = Cumulative Illness Rating Scale - Geriatric, BPRS = Brief Psychiatric Rating Scale, SAS = Simpson-Angus Scale, QLS= Quality of Life Scale, SANS = Scale for the Assessment of Negative Symptoms |
||||
Statistical preprocessing. All statistical tests were run with R version 3.5.3. We [WILL] regressed age, sex, and a DWI scan quality composite score from the \(X\) set, and age and sex from the social cognition scores in the \(Y\) set, and used neurocognition scores normed for age and sex. For ease of interpretation, all variables in the \(X\) and \(Y\) sets were standardized via Z-scoring. Though variable sets were not found to be univariate and multivariate normal, their observed distributions, in conjunction with sample size, were sufficient to assume properties of the central limit theorum were in effect and thus assumptions of CCA were satisfied. We also examined raw correlations within and between the \(X\) and \(Y\) sets, to affirm our conceptual grouping. As expected, we found small-to-moderate positive correlations between brain variables (\(R_{xx}\) matrix mean r=.21, range=-.40-.79) and moderate-to-strong correlations between neurocognitive and social cognitive variables (\(R_{yy}\) matrix mean r=.38, range=-.04-.69; see Supplementary Figure 1). We also found that several variables in the \(R_{xy}\) matrix showed some multicolinearity, as visualized by an adjusted \(R_{xy\omega}\) matrix (see Supplementary Figure 2).
Selection of canonical variates. The CCA analysis produced 14 canonical functions (analogous with the size of the smaller \(Y\) set). First, we employed a permutation test (500 bootstraps) to evaluate the null hypothesis of no correlation between the \(X\) and \(Y\) sets. Wilk’s lambda suggested no correlation between the sets (p=.35). Similar values were seen when employing the Hotelling-Lawley Trace (p=.232), the Pillai-Bartlett Trace (p=.476), and Roy’s Largest Root (p=.052), with all canonical correlations included in all models. We proceeded to review the functions stepwise for significance, magnitude, and redundancy (Hair, Anderson, and Tatham, 1998). Using Wilk’s lambda, we found no functions were significant; the first function – which by definition falls closest to significance – found \(\lambda\)=<.00, F(1008, 1112.5)=1.02, p=.360. The Stewart-Love redundancy index (i.e., variance in one set that can be explained by the other set) was 1% for the \(X\) set and 0.5% for the \(Y\) set. Though they should not be interpreted due to lack of significance, Table 2 shows standardized canonical function coefficients (\(\beta\)), structure coefficents (\(r_s\)), and squared structure coefficients \(r_s^2\) (here analogous with communalities) for the first derived function. Canonical weights (\(\beta\)) are included for illustrative purposes, and should not be interpreted even if the model were significant, as they are unstable in the face of multicollinearity, which we have already demonstrated is a feature of the \(R_{XX}\) matrix in Supplementary Figure 2. Interestingly, there was a very strong correlation, \(R_c\)=.866, between participants’ CCA scores in the \(X\) and \(Y\) for the first function, albeit insignificant (see Figure 2 for a visualization of the association).
Table 2.
| Variable | \(\beta\) | \(r_s\) | \(r_s^2\) (\(h^2\)) |
|---|---|---|---|
| X | |||
| AF (R) | -0.144 | 0.02 | 0.000 |
| AF (L) | 0.137 | -0.073 | 0.005 |
| CB (R) | 0.090 | -0.22 | 0.049 |
| CB (L) | -0.169 | -0.165 | 0.027 |
| CC1 (C) | -0.086 | -0.116 | 0.014 |
| CC2 (C) | 0.089 | -0.14 | 0.020 |
| CC3 (C) | -0.191 | -0.102 | 0.010 |
| CC4 (C) | -0.086 | -0.029 | 0.001 |
| CC5 (C) | 0.372 | 0.148 | 0.022 |
| CC6 (C) | -0.008 | -0.033 | 0.001 |
| CC7 (C) | 0.079 | -0.007 | 0.000 |
| CPC (L) | 0.302 | 0.084 | 0.007 |
| CPC (R) | -0.259 | -0.316 | 0.100 |
| CR.F (L) | 0.309 | 0.075 | 0.006 |
| CR.F (R) | 0.154 | 0.006 | 0.000 |
| CR.P (R) | 0.048 | -0.028 | 0.001 |
| CR.P (L) | -0.045 | 0.018 | 0.000 |
| CST (L) | 0.298 | 0.009 | 0.000 |
| CST (R) | -0.065 | -0.054 | 0.003 |
| EC (R) | 0.005 | -0.081 | 0.007 |
| EC (L) | 0.015 | -0.067 | 0.005 |
| EmC (L) | -0.020 | 0.006 | 0.000 |
| EmC (R) | 0.219 | 0.037 | 0.001 |
| ICP (L) | -0.274 | -0.312 | 0.097 |
| ICP (R) | 0.020 | -0.145 | 0.021 |
| ILF (R) | 0.136 | 0.127 | 0.016 |
| ILF (L) | -0.165 | -0.026 | 0.001 |
| Intra.CBLM.I.P (R) | -0.318 | -0.118 | 0.014 |
| Intra.CBLM.I.P (L) | 0.070 | 0.029 | 0.001 |
| Intra.CBLM.PaT (L) | 0.127 | 0.028 | 0.001 |
| Intra.CBLM.PaT (R) | -0.284 | -0.156 | 0.024 |
| IOFF (L) | -0.144 | -0.073 | 0.005 |
| IOFF (R) | 0.093 | 0.185 | 0.034 |
| MCP (C) | 0.022 | -0.009 | 0.000 |
| MdLF (R) | -0.352 | 0.028 | 0.001 |
| MdLF (L) | 0.125 | -0.061 | 0.004 |
| PLIC (R) | 0.328 | -0.029 | 0.001 |
| PLIC (L) | -0.192 | -0.211 | 0.044 |
| SF (R) | 0.186 | -0.087 | 0.008 |
| SF (L) | -0.202 | -0.199 | 0.040 |
| SLF.II (L) | 0.318 | 0.184 | 0.034 |
| SLF.II (R) | -0.137 | -0.064 | 0.004 |
| SLF.III (R) | -0.249 | -0.193 | 0.037 |
| SLF.III (L) | -0.114 | -0.1 | 0.010 |
| SO (L) | -0.211 | -0.071 | 0.005 |
| SO (R) | 0.021 | 0.085 | 0.007 |
| SP (R) | -0.148 | -0.13 | 0.017 |
| SP (L) | -0.279 | -0.19 | 0.036 |
| Sup.F (R) | -0.170 | -0.09 | 0.008 |
| Sup.F (L) | 0.254 | -0.148 | 0.022 |
| Sup.FP (L) | -0.196 | -0.112 | 0.012 |
| Sup.FP (R) | -0.061 | -0.187 | 0.035 |
| Sup.O (L) | -0.039 | -0.039 | 0.002 |
| Sup.O (R) | -0.098 | 0.074 | 0.006 |
| Sup.OT (R) | 0.205 | 0.237 | 0.056 |
| Sup.OT (L) | 0.017 | -0.109 | 0.012 |
| Sup.P (L) | -0.001 | -0.05 | 0.003 |
| Sup.P (R) | 0.382 | 0.023 | 0.001 |
| Sup.PO (L) | -0.321 | -0.257 | 0.066 |
| Sup.PO (R) | 0.072 | 0.057 | 0.003 |
| Sup.PT (R) | 0.187 | 0.123 | 0.015 |
| Sup.PT (L) | -0.010 | 0.059 | 0.003 |
| Sup.T (R) | 0.094 | 0.08 | 0.006 |
| Sup.T (L) | 0.048 | 0.052 | 0.003 |
| TF (L) | -0.462 | -0.073 | 0.005 |
| TF (R) | 0.014 | -0.106 | 0.011 |
| TO (L) | 0.174 | 0.053 | 0.003 |
| TO (R) | -0.021 | -0.106 | 0.011 |
| TP (L) | 0.051 | 0.022 | 0.000 |
| TP (R) | -0.022 | 0.001 | 0.000 |
| UF (L) | 0.339 | 0.069 | 0.005 |
| UF (R) | -0.327 | -0.209 | 0.044 |
| Y | |||
| Processing speed | -0.475 | -0.461 | 0.212 |
| Attention & vigilance | -0.459 | -0.482 | 0.233 |
| Working memory | -0.297 | -0.286 | 0.082 |
| Verbal learning | -0.181 | -0.352 | 0.124 |
| Visual learning | -0.181 | -0.423 | 0.179 |
| Problem solving | 0.359 | -0.152 | 0.023 |
| RMET | 0.217 | -0.013 | 0.000 |
| RAD | 0.025 | -0.143 | 0.020 |
| ER-40 | 0.470 | 0.086 | 0.007 |
| TASIT 1 | -0.267 | -0.164 | 0.027 |
| TASIT 2 | -0.563 | -0.265 | 0.070 |
| TASIT 3 | 0.698 | 0.089 | 0.008 |
| IRI | 0.055 | 0.201 | 0.040 |
| EA | 0.464 | 0.192 | 0.037 |
| Note: | |||
|
\(r_s\) values > .450 are emphasized, following convention; \(\beta\) = standardized canonical function coefficient; \(r_s\) = structure coefficient; \(r_s^2\) = squared structure coefficient, here also communality \(h^2\) |
|||
Validation. Lastly, we validated our model via iterative feature removal, i.e., we removed one of features from the combined p=86 of the \(X\) and \(Y\) sets, ran the CCA, and compared the derived canonical correlation coefficients. This procedure leveraged the CCA property that each variable relates to all other variables in both sets: if the model is stable, canonical correlation estimates will remain similar; if unstable, estimate will vary widely. Values were similar across all iterations, suggesting that our model is stable (Supplementary Figure 3).
Exploratory analysis. Though increasing evidence suggests that neurocognition and social cognition are dimensional constructs on which HC and SSD fall along a continuum, as opposed to discrete classes, we were curious to verify that HC and SSD participants exhibited a complementary pattern of correlation within or between \(X\) and \(Y\) sets. We observed that their respective \(R_{xx}\), \(R_{yy}\), and \(R_{xy}\) matrices are very similar across HC and SSD populations.
Discussion
Summary. We performed a CCA on a large \(X\) set of white matter tracts, and a small \(Y\) set of neurocognitive and social cognitive variables, across 162 participants along the HC-SSD spectrum. CCA no significant relationship between these sets, though CCA scores on the first canonical function were highly correlated, \(R_c\)=.866. Thus, it appears that the structural integrity of white matter tracts is generally predictive of neurocognitive and social cognitive ability. However, evaluating the strength of correlations between \(X\) and \(Y\) sets suggests that no particular variables within each set contributed uniquely to the correlation. This may be an upshot of other features of the data distributions (e.g., non-normality), and/or derivation (e.g., pre-processing).
The null results of the present analysis are also a consequence of the number of features included in the \(X\) set. By way of contrast, we ran an otherwise comparable analysis with just six brain features (bilateral AF, UF, and ILF) and found one significant component, λ=.397, F(48, 382.93)=1.65, p=.006, with a large magnitude, \(R_c\)=.571. In that analysis, examination of structure coefficients (\(r_s\)) (which reveal the correlation between a given variable and the synthetic set to which it belongs, irrespective of the contribution of other variables) for the \(X\) set revealed that FA values in the right ILF contributed most highly (\(r_s\)=.832), followed by the left ILF (\(r_s\)=.509) and right AF (\(r_s\)=.464). In contrast, the left AF and right UF made modest contributions, and the left UF a negligible contribution, suggesting these tracts may not be very strongly related to the synthetic variable combining neurocognition and social cognition. In that analysis, examination of the \(Y\) set also showed some very high structure coefficients. Specifically, the social cognition variables (in that analysis, represented by Simulation (\(r_s\)=.911) and Mentalizing (\(r_s\)=.943) factor scores; Oliver, 2018), made substantive contributions. Additionally, four of six neurocognition variables made large contributions: Processing speed (\(r_s\)=.593), Working memory (\(r_s\)=.512), Verbal learning (\(r_s\)=.547), and Visual learning (\(r_s\)=.489). All of the \(Y\) set structure coefficients had the same sign, indicating that all variables are positively related. The Attention & vigilance and Problem solving factors did not reveal high structure coefficients, suggesting that those factors were not strongly related to white matter integrity in the included tracts, and also provided evidence that neurocognition has multidimensional representation in the brain, as well as behaviour.
Conclusion
We used the multivariate technique CCA to uncover links between integrity of white matter microstructure (operationalized as FA) and domains of neurocognition and social cognition. To the best of our knowledge, this is the first study to examine how tracts implicated in social cognition may also subserve select facets of neurocognition, in a large sample of SSD and HCs.
At least one aspect of CCA makes it particularly attractive to the task of elucidating brain and behaviour links in psychiatry: it holistically integrates observations from transdiagnostic groups (here, HC and SSD), and thus avoids the increasingly noted pitfalls of case-control design. Accumulating evidence suggests that the neurobiology underlying particular deficits in SSD is also evident in HC. Our results suggest that this relationship holds for the association between white matter microstructure and neurocognitive and social cognitive deficits in both HC and SSD. Future research should further test this association transdiagnostically, for example, by including data from other disorders with noted neurocognitive and social cognitive deficits.
Several limitations should be noted, both of the present study design and the CCA statistical method. In relation to the former, perhaps the greatest limitation is that our current sample size (limited by use of only data from the PRISMA scanners, until we implement harmonization) is insufficient to be confident in our conclusions. Relatedly, obtaining an independent replication sample is essential for cross-validation, i.e., the ability to assess the model’s ability to generalize to unseen, unrelated samples. Pertaining to methods, it must be remembered that as a linear model, CCA assumes additive covariation patterns, which may prove a simplification of brain-behaviour relationships. Finally, CCA is ‘merely’ correlational; investigating the cause(s) of the reported correlations remains a challenging issue for future experimental work.
Nonetheless, multivariate analyses that illuminate the shared and distinct neural basis of disabling deficits, such as neurocognitive and social cognitive impairment, are sure to accelerate the design of novel, biologically-targeted treatments for SSD, and psychiatry more broadly.
Code availability. All analysis code is available on GitHub (private repo until code review): github.com/navonacalarco/thesis/tree/master/SPINS/analyses
Relevant sub-analyses.
(1) Summary of psychotropic medications
(2) Calculation of CPZE
(3) Automated DWI QC procedure – description
(4) Automated DWI QC procedure – results
(5) Review of data distribution – missing values – \(X\) set
(6) Review of data distribution – univariate outliers – \(Y\) set
(7) Review of multivariate and univariate normality – \(X\) set
(8) Review of multivariate and univariate normality – \(Y\) set
(9) Multivariate interpolation of missing data – \(X\) set
(10) Multivariate interpolation of missing data – \(Y\) set
(11) Review of site effect in prospectively harmonized data
(12) RISH harmonization: participant matching
(13) RISH harmonization: participant characteristics
(14) RISH harmonization: problem with brain extraction, requiring new preprocessing
(15) Comparable analysis with smaller p x n for statistics class
References
Behdinan, T., Foussias, G., Wheeler, A.L., Stefanik, L., Felsky, D., Remington, G., Rajji, T.K., Chakravarty, M.M., Voineskos, A.N. (2015). Neuroimaging predictors of functional outcomes in schizophrenia at baseline and 6-month follow up. Schizophrenia Research., 169(1), 69-75.
Hair, J.F Jr., Anderson, R.E., Tatham. R.L., Black, W.C. (1998). Chapter 8. Multivariate Data Analysis, 5th edition. Prentice Hall.
Oliver, L.D., Haltigan, J.D., Gold, J.M., Foussias, G., DeRosse, P., Buchanan, R.W., Malhotra, A.K., Voineskos, A.N. (2019). Lower- and higher-level social cognitive factors across individuals with Schizophrenia Spectrum Disorders and healthy controls: Relationship with neurocognition and functional outcome. Schizophrenia Bulletin, 45(3), 629-638.
van Buuren, S. (2019). mice: Multivariate Imputation by Chained Equations . R package veersion 3.7.0. https://cran.r-project.org/web/packages/mice/mice.pdf
Voineskos, A.N., Foussias, G., Lerch, J., Felsky, D., Remington, G., Rajji, T.K., Lobaugh, N., Pollock, B.G., Mulsant, B.H. (2013). Neuroimaging evidence for the deficit subtype of schizophrenia. JAMA Psychiatry, 70(5), 472-480.
Wheeler, A.L., Wessa, M., Szeszko, P.R., Foussias, G., Chakravarty, M.M., Lerch, J.P., DeRosse, P., Remington, G., Mulsant, B.H., Linke, J., Malhotra, A.K., Voineskos, A.N. (2015). Further Neuroimaging Evidence for the Deficit Subtype of Schizophrenia: A Cortical Connectomics Analysis. JAMA Psychiatry, 72(5), 446-455.
Zhang, F., Wu, Y., Norton, I., Rathi, Y., Makris, N., O’Donnell, L.J. (2018). An anatomically curated fiber clustering white matter atlas for consistent white matter tract parcellation across the lifespan. NeuroImage, 179, 429-447.
Supplementary Tables
Supplementary Table 1. List of the 41 unique (74 combinations) of white matter tracts parcellated by the Slicer ORG atlas.| Abbreviation | Full name | Hemisphere | |
|---|---|---|---|
| Association tracts | |||
| 1 | AF | arcuate fasciculus | LR |
| 2 | CB | cingulum bundle | LR |
| 3 | EC | external capsule | LR |
| 4 | EmC | extreme capsule | LR |
| 5 | ILF | inferior longitudinal fasciculus | LR |
| 6 | IoFF | inferior occipito-frontal fasciculus | LR |
| 7 | MdLF | middle longitudinal fasciculus | LR |
| 8 | PLIC | posterior limb of internal capsule | LR |
| 9 | SLF I | superior longitudinal fasciculus I | LR |
| 10 | SLF II | superior longitudinal fasciculus II | LR |
| 11 | SLF III | superior longitudinal fasciculus III | LR |
| 12 | UF | uncinate fasciculus | LR |
| Cerebellar tracts | |||
| 13 | CPC | cortico-ponto-cerebellar | LR |
| 14 | ICP | inferior cerebellar peduncle | LR |
| 15 | Intra-CBLM-I&P | intracerebellar input and Purkinje tract | LR |
| 16 | Intra-CBLM-PaT | intracerebellar parallel tract | LR |
| 17 | MCP | middle cerebellar peduncle | C |
| Commissural tracts | |||
| 18 | CC 1 | corpus callosum 1 | C |
| 19 | CC 2 | corpus callosum 2 | C |
| 20 | CC 3 | corpus callosum 3 | C |
| 21 | CC 4 | corpus callosum 4 | C |
| 22 | CC 5 | corpus callosum 5 | C |
| 23 | CC 6 | corpus callosum 6 | C |
| 24 | CC 7 | corpus callosum 7 | C |
| Projection tracts | |||
| 25 | CST | corticospinal tract | LR |
| 26 | CR-F | corona-radiata-frontal | LR |
| 27 | CR-P | corona-radiata-parietal | LR |
| 28 | SF | striato-frontal | LR |
| 29 | SO | striato-occipital | LR |
| 30 | SP | striato-parietal | LR |
| 31 | TF | thalamo-frontal | LR |
| 32 | TO | thalamo-occipital | LR |
| 33 | TP | thalamo-parietal | LR |
| Superficial tracts | |||
| 34 | Sup-F | superficial-frontal | LR |
| 35 | Sup-FP | superficial-frontal-parietal | LR |
| 36 | Sup-O | superficial-occipital | LR |
| 37 | Sup-OT | superficial-occipital-temporal | LR |
| 38 | Sup-P | superficial-parietal | LR |
| 39 | Sup-PO | superficial-parietal-occipital | LR |
| 40 | Sup-PT | superficial-parietal-temporal | LR |
| 41 | Sup-T | superficial-temporal | LR |
| Note: | |||
| LR = left and right, C = commissural | |||
Supplementary Figures
Supplementary Figure 1. Correlation matrices for the \(X\) and \(Y\) sets. A. The \(R_{xx}\) matrix (white matter FA) generally shows small-to-moderate positive correlations between tracts. The order of tracts is alphabetical along the axes, as captured in Supplementary Table 1. B. The \(R_{yy}\) matrix (neurocognition and social cognition variables) shows moderate-to-strong positive correlations between all variables.
Supplementary Figure 2. Cross-correlation, omega, and difference matrices. A. The \(R_{xy}\) matrix shows cross-correlations between variables in the \(X\) and \(Y\) sets. We see some evidence of multicolinearity. B. The omega matrix shows the \(R_{xy}\) matrix adjusted for redundancy. The omega matrix is calculated as the product of the inverse of the Choleski factorization of the \(R_{xx}\) matrix, the \(R_{xy}\), and the inverse of the Choleski factorization of the \(R_{xy}\) matrix; C. The difference matrix shows the magnitude of difference between the original \(R_{xy}\) and difference matrices.
Supplementary Figure 3. Model validation. We validated our model using iterative feature removal, i.e., we removed one of 14 features from the combined \(X\) and \(Y\) sets, ran the CCA, and compared the derived canonical correlation coefficients. The plot shows the 86 comparisons (black dots), as well the original model (red dots), across all 14 canonical functions. We see that, though iteratively removing a variable does alter the canonical correlation coefficients, the change is not drastic, nor statistically significant.