Comparison of Functional Connectivity Estimated from Concatenated Task-State Data from Block-Design Paradigm with That of Continuous Task

Abstract

Functional connectivity (FC) analysis with data collected as continuous tasks and activation analysis using data from block-design paradigms are two main methods to investigate the task-induced brain activation. If the concatenated data of task blocks extracted from the block-design paradigm could provide equivalent FC information to that derived from continuous task data, it would shorten the data collection time and simplify experimental procedures, and the already collected data of block-design paradigms could be reanalyzed from the perspective of FC. Despite being used in many studies, such a hypothesis of equivalence has not yet been tested from multiple perspectives. In this study, we collected fMRI blood-oxygen-level-dependent signals from 24 healthy subjects during a continuous task session as well as in block-design task sessions. We compared concatenated task blocks and continuous task data in terms of region of interest- (ROI-) based FC, seed-based FC, and brain network topology during a short motor task. According to our results, the concatenated data was not significantly different from the continuous data in multiple aspects, indicating the potential of using concatenated data to estimate task-state FC in short motor tasks. However, even under appropriate experimental conditions, the interpretation of FC results based on concatenated data should be cautious and take the influence due to inherent information loss during concatenation into account.

Figure 1

A schematic diagram for extracting task blocks from block-design session. A temporal shift of 2 volumes was adopted for the hemodynamic delay.

Figure 2

Scatter plots of FC strength for continuous segment 4 and concatenated data for a typical subject in LHG (a) and a typical subject in RHG (b). There are a total of 6670 FCs between 116 nodes (i.e., ROIs), and each circle represents one FC.

Figure 3

The similarities of association matrices (r-to-z transformed) in comparison pair as well as those in reference pair in LHG and RHG for four continuous segments. Error bars denote the standard deviation across subjects.

Figure 4

The overlap ratios of adjacency matrices for comparison pair as well as those for reference pair under sparsity from 0.1 to 0.5 with an increment of 0.02 in LHG (a) and RHG (b) for continuous segment 4. Error bars denote the standard deviation across subjects. Asterisk (∗) denotes that the overlap ratio for reference pair is significantly greater (FDR corrected, p < 0.05) than that for comparison pair under the sparsity.

Figure 5

The normalized CC and Lambda of continuous segment 4 and concatenated data in LHG ((a) and (c)) and RHG ((b) and (d)), respectively. Error bars denote the standard deviation across subjects.

Figure 6

Group-level seed-based FC maps of continuous segment 4 ((a) and (c)) and concatenated data ((b) and (d)) in LHG ((a) and (b)) and RHG ((c) and (d)). The index of color bar indicates t value of t-test. Voxels with hot color have positive FC with corresponding seed region (x = ±38, y = −26, and z = 56), indicated by a green sphere, with cool color for negative FC. R values and p values ((a) versus (b) and (c) versus (d)) indicate that the similarities between the seed-based FC maps of continuous segment 4 and concatenated data are significant.

Figure 7

The similarities of seed-based FC maps (r-to-z transformed) in comparison pair as well as those in reference pair in LHG and RHG for four continuous segments. Error bars denote standard deviation across subjects.