18-08-05¶

Linked dimensions of psychopathology and connectivity in functional brain networks

Xia, C. H., Ma, Z., Ciric, R., Gu, S., Betzel, R. F., Kaczkurkin, A. N., ... & Cui, Z. (2018). Linked dimensions of psychopathology and connectivity in functional brain networks. Nature Communications, 9(1), 3003.

Original | Mendeley | Github

Overview¶

1. Intro
2. Results
   • 2.1 Linked dimensions of psychopathology and connectivity
   • 2.2 Brain-guided dimensions of psychopathology cross clinical diagnostic categories
   • 2.3 Common and dissociable patterns of connectivity
   • 2.4 Developmental effects and sex differences
   • 2.5 Linked dimensions are replicated in an independent sample
3. Discussion
4. Methods
   • 4.1 Participants
   • 4.2 Psychiatric assessment
   • 4.3 Image acquisition
   • 4.4 Structural preprocessing
   • 4.5 Functional preprocessing
   • 4.6 Network construction
   • 4.7 Dimensionality reduction
   • 4.8 Sparse canonical correlation analysis
   • 4.9 Grid search for regularization parameters
   • 4.10 Permutation testing
   • 4.11 Resampling procedure
   • 4.12 Network module analysis
   • 4.13 Analysis of common connectivity features across dimensions
   • 4.14 Analysis of age effects and sex differences

1. Intro¶

Insel 2015

Bassett 2017

module-specific:

Gordon 2014

Bassett 2018

between modules

Satterthwaite 2013

"neurodeve-lopmental connectopathies" Paus 2005

limitations: 1. case-ctrl approach 2. driven by cov in the clinical symptomatolgy, rather than brain & behavior feat 3. youth studies: small samples Bzdok 2017

sCCA Witten 2009

Drysdale 2017

2. Result¶

2.1 Linked dimensions of psychopathology and connectivity¶

2.2 Brain-guided dimensions of psychopathology cross clinical diagnostic categories¶

2.3 Common and dissociable patterns of connectivity¶

2.4 Developmental effects and sex differences¶

2.5 Linked dimensions are replicated in an independent sample¶

3. Discussion¶

4. Online methods¶

4.1 Participants¶

4.2 Psychiatric assessment¶

4.3 Image acq¶

4.4 Structural prep¶

4.5 Functional prep¶

1. magnetic field inhomogeneities
2. 4 initial vol
3. realignmnet
4. intensity outliers
5. demeaning & removal of any linear / quatdric trends

4.6 Network construction¶

4.7 Dimensinality reduction¶

MAD (Median Absolute Deviation)

Def: MAD

$(| \mathbf{X}_i - \text{median}(\mathbf{X})|)$

4.8 sCCA¶

• $\mathbf{X}_{n \times p}$, $\mathbf{Y}_{n \times q}$: mat
• $n$: num of obs (participants)
• $p$, $q$: num of var (feat)
• $\mathbf{u}$, $\mathbf{v}$
• $c_1$, $c_2$: constant

$$\begin{align*} \tag{1} \text{maximize } \mathbf{u}^T\mathbf{X}^T\mathbf{Yv}, \\ \text{subject to } ||\mathbf{u}||^2_2 ≤ 1, ||\mathbf{v}||^2_2 ≤ 1, \\ ||\mathbf{u}||_1 ≤ c_1, ||\mathbf{v}||_1 ≤ c_2 \end{align*}$$

$$\begin{align*} L_1\text{-norm:} \quad ||\mathbf{x}||_1 &:= \sum_{i=1}^n |x_i| \\ L_2\text{-norm:} \quad ||\mathbf{x}||_2 &:= \sqrt{ \sum_{i=1}^n{x_i^2}} \end{align*}$$

4.9 Grid search for regularization parameters¶

4.10 Permutation testing¶

4.11 Resampling procedure¶

4.12 Network module analysis¶

4.13 Analysis of common connectivity features across dimensions¶

4.14 Analysis of age effects and sex differences¶