Network-based statistics distinguish anomic and Broca’s aphasia

Abstract

Introduction

Aphasia is a speech-language impairment commonly caused by damage to the left hemisphere. The neural mechanisms that underpin different types of aphasia and their symptoms are still not fully understood. This study aims to identify differences in resting-state functional connectivity between anomic and Broca’s aphasia measured through resting-state functional magnetic resonance imaging (rs-fMRI).

Methods

We used the network-based statistic (NBS) method, as well as voxel- and connectome-based lesion symptom mapping (V-, CLSM), to identify distinct neural correlates of the anomic and Broca’s groups. To control for lesion effect, we included lesion volume as a covariate in both the NBS method and LSM.

Results

NBS identified a subnetwork located in the dorsal language stream bilaterally, including supramarginal gyrus, primary sensory, motor, and auditory cortices, and insula. The connections in the subnetwork were weaker in the Broca’s group than the anomic group. The properties of the subnetwork were examined through complex network measures, which indicated that regions in right inferior frontal sulcus, right paracentral lobule, and bilateral superior temporal gyrus exhibit intensive interaction. Left superior temporal gyrus, right postcentral gyrus, and left supramarginal gyrus play an important role in information flow and overall communication efficiency. Disruption of this network underlies the constellation of symptoms associated with Broca’s aphasia. Whole-brain CLSM did not detect any significant connections, suggesting an advantage of NBS when thousands of connections are considered. However, CLSM identified connections that differentiated Broca’s from anomic aphasia when analysis was restricted to a hypothesized network of interest.

Discussion

We identified novel signatures of resting-state brain network differences between groups of individuals with anomic and Broca’s aphasia. We identified a subnetwork of connections that statistically differentiated the resting-state brain networks of the two groups, in comparison with standard CLSM results that yielded isolated connections. Network-level analyses are useful tools for the investigation of the neural correlates of language deficits post-stroke.

Appendix. Performance evaluation of NBS

For baseline comparison with the NBS method, mass univariate permutation testing with multiple comparison was performed. The p-value for comparing the jth edge weight was calculated by $Pr\left( |t^{*}_j| \ge |t_j|\right)$, where $t^{*}_j$ and $t_j$ were the test statistics computed from permuted data and original data respectively. This p-value describes how likely the absolute value of test statistic from permuted data exceeds the absolute value of test statistic from original data. Then, the following multiple comparison procedures were applied to correct multiple p-values and control the FWER: Bonferroni correction (Bonferroni 1936), Holm’s Bonferroni correction (Holm 1979), and false discovery rate (FDR) control (Benjamini and Hochberg 1995).

We used three simulation studies to evaluate the empirical performance of the NBS method against mass univariate testing with multiple comparison. We assessed two aspects of the performance: (1) sensitivity or true positive rate (TPR): the proportion of connections or edges containing group differences that are correctly identified; (2) $1-$ specificity or false positive rate (FPR): the proportion of edges without differences that are misclassified. Ideally, TPR $=1$ (all edges that differ between the two groups are identified), and FPR $=0$ (all edges that do not differ between the two groups are not identified). Suppose H is the set of edges that differ between the groups, R is the set of edges that do not differ between the groups, and $\hat{h}$ is the set of edges comprising the subnetwork identified by a specific method (NBS, baseline mass univariate testing with multiple comparison). The TPR was then calculated by $|H\cap \hat{h}|/|H|$ and the FPR by $|R\cap \hat{h}|/|R|$.

Fig. 9

Simulation setup: The white blocks in the left side’s matrices represent contrast edges, whose weights differ between the two groups. The white blocks in the right side’s matrices are the subnetwork or connected component identified by the NBS method or the edges declared significant by multiple comparison

In each of the three studies, we generated two groups of p-node networks. The group sizes are $n_1=n_2=10$ for all three studies. In each network, the weight of the edge between node i and j was generated by $w_{ij} \sim N(arctanh(r_{ij}), \sigma _{w}^2)$ with $r_{ij} \sim U(-1,1)$ and $\sigma _{w} = 1/\sqrt{p(p-1)/2-3}$. We randomly chose $C_r \%$ of $p(p-1)/2$ edges to differ in weights between Group 1 and 2, and refer to these edges as contrast edges. The weights of contrast edges were generated independently with

$$\begin{aligned} w_{ij}^{*}=\left\{ \begin{array}{ll} w_{ij} + w^*, \text {for } w_{ij} \ge 0\\ w_{ij} - w^*, \text {for } w_{ij} < 0 \text {,} \end{array} \right. \end{aligned}$$

where $w^{*}\sim N(0.03, 0.01)$. We compared the performance of the methods via different p and $C_r$ values in the three studies.

Study 1.:

$p=20$ and $C_r \% = 10\%$.

Study 2.:

$p=40$ and $C_r \% = 10\%$.

Study 3.:

$p=40$ and $C_r \% = 5\%$.

After the networks were generated, NBS with threshold 2.5 and mass univariate testing with multiple comparison were performed (Fig. 9). We repeated the simulation process 5,000 times for each study. Average TPR and FPP were computed respectively.

Table 3 The average TPR and FPR of the NBS with threshold 2.5 and mass univariate testing with multiple comparison, in 5,000 simulations

Simulation results

Table 3 summarizes results of the three simulation studies. In Study 1, the NBS method has the largest TPR and FPR, while the mass univariate testing with FDR detects a desirable proportion of contrast edges and contains a small number of false discoveries. In Study 2, as the network size expands, the TPRs by the NBS and mass univariate testing methods increase, whereas the FPRs increase by the NBS and FDR methods and decrease by the Bonferroni and Holm’s Bonferroni methods. In Study 3, when fewer contrast edges are placed in either of the two groups, the TPRs by the NBS and FDR methods decrease and the FPRs by the NBS and the two Bonferroni corrections stay similar to Study 2.

In summary, compared with mass univariate testing, the NBS method detects small group differences well under various network sizes and proportions of contrast edges. For the mass univariate testing, the FDR has the highest power and a favorable FPR; Bonferroni and Holm’s Bonferroni corrections are highly conservative in detecting contrast edges. Additionally, we find that the computation speed is mainly affected by network size.