Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2012 Jan;22(1):158-65.
doi: 10.1093/cercor/bhr099. Epub 2011 May 26.

Decoding subject-driven cognitive states with whole-brain connectivity patterns

W R Shirer  1 , S Ryali, E Rykhlevskaia, V Menon, M D Greicius
Affiliations

Decoding subject-driven cognitive states with whole-brain connectivity patterns

W R Shirer et al. Cereb Cortex. 2012 Jan.

Abstract

Decoding specific cognitive states from brain activity constitutes a major goal of neuroscience. Previous studies of brain-state classification have focused largely on decoding brief, discrete events and have required the timing of these events to be known. To date, methods for decoding more continuous and purely subject-driven cognitive states have not been available. Here, we demonstrate that free-streaming subject-driven cognitive states can be decoded using a novel whole-brain functional connectivity analysis. Ninety functional regions of interest (ROIs) were defined across 14 large-scale resting-state brain networks to generate a 3960 cell matrix reflecting whole-brain connectivity. We trained a classifier to identify specific patterns of whole-brain connectivity as subjects rested quietly, remembered the events of their day, subtracted numbers, or (silently) sang lyrics. In a leave-one-out cross-validation, the classifier identified these 4 cognitive states with 84% accuracy. More critically, the classifier achieved 85% accuracy when identifying these states in a second, independent cohort of subjects. Classification accuracy remained high with imaging runs as short as 30-60 s. At all temporal intervals assessed, the 90 functionally defined ROIs outperformed a set of 112 commonly used structural ROIs in classifying cognitive states. This approach should enable decoding a myriad of subject-driven cognitive states from brief imaging data samples.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
Functional parcellation of the brain into 90 regions of interest that cover the majority of cortical and subcortical gray matter. Group ICA applied to the resting-state data of 15 subjects yielded 14 ICNs of which 5 are shown in A (for all 14 ICNs, see Supplementary Fig. S2). Each ICN is thresholded to generate between 2 and 12 ROIs per ICN. When all 90 ROIs across the 14 ICNs are overlaid on a single brain image (B) the majority of cortical and subcortical gray matter is covered.
Figure 2.
Figure 2.
Subject-driven episodic memory recall drives changes in whole-brain functional connectivity. A single subject's connectivity matrix is shown for the rest scan in A. Cells colored in yellow-red indicate a positive pairwise correlation between 2 ROIs; green-blue cells indicate negative pairwise correlations. Coarse anatomic labels for each ICN are indicated along the x- and y-axes (more detailed anatomic information is available in Supplementary Table S1). Each ICN is bracketed by black bars and divided into 2–12 ROIs. The strong within-ICN correlations are evident along the diagonal. The same subject's memory state connectivity matrix is shown in B. Subtracting the rest state matrix from the memory state matrix provides the difference matrix shown in C where connectivity within the RSC/MTL network is shown to increase during the memory task. A paired-sample t-test of the state matrices across all 14 subjects (D) reveals changes in connectivity both between and within ICNs. These within-ICN changes (orange arrow) can also be detected by performing a paired-sample t-test on the individual subject ICA data (E). This analysis reveals clusters in the RSC/MTL network whose connectivity increases significantly during the memory scan compared with the rest scan.
Figure 3.
Figure 3.
Distinct across-subject patterns of whole-brain connectivity for 4 subject-driven cognitive states. For each of the 4 states, cells of interest which showed significant state-specific positive or negative correlations were included in the group-level state matrix. These state matrices are shown in A–D. The orange arrow in B indicates strong connectivity within the RSC/MTL network in the group-level memory state matrix. In the subtraction task (D), connectivity within the IPS/Prefrontal ICN is increased (blue arrow) but the classification algorithm also highlights increased connectivity between this ICN and the basal ganglia ICN (green arrow).
Figure 4.
Figure 4.
Classification accuracy remains high with scans as short as 1 min. The classification algorithm was tested initially on the full 10-min scans but then on increasingly shorter scan lengths. In each case, the shorter scan lengths are taken from the beginning of the scan (i.e. 0.5 min refers to the first 30 s of the scan). Eleven different scan lengths from 30 s to 10 min were evaluated. The orange line refers to the overall accuracy in distinguishing all 4 states. Accuracy for individual states is shown in the other 4 colors. An accuracy of 25% reflects chance level classification. The overall accuracy remains at 80% with just 1 min of data. With scan lengths below 1 min, overall accuracy tends to decrease, though all 4 scans were identified with significant accuracy with only 30 s of data (P < 0.001).
Figure 5.
Figure 5.
Functional ROIs outperform structural ROIs. We performed classification with 112 structural ROIs from the Automated Anatomical Labeling (AAL) Atlas and 90 functional ROIs identified by ICA on resting-state data from an independent sample. Classification was performed with both sets of ROIs at 11 different scan lengths. In all comparisons, classification with functional ROIs was substantially more accurate than classification with the AAL Atlas ROIs.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Beckmann CF, DeLuca M, Devlin JT, Smith SM. Investigations into resting-state connectivity using independent component analysis. Philos Trans R Soc Lond B Biol Sci. 2005;360:1001–1013. - PMC - PubMed
    1. Bellec P, Perlbarg V, Jbabdi S, Pelegrini-Issac M, Anton JL, Doyon J, Benali H. Identification of large-scale networks in the brain using fMRI. Neuroimage. 2006;29:1231–1243. - PubMed
    1. Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med. 1995;34:537–541. - PubMed
    1. Chadwick MJ, Hassabis D, Weiskopf N, Maguire EA. Decoding individual episodic memory traces in the human hippocampus. Curr Biol. 2010;20:544–547. - PMC - PubMed
    1. Chang C, Glover GH. Effects of model-based physiological noise correction on default mode network anti-correlations and correlations. Neuroimage. 2009;47:1448–1459. - PMC - PubMed

Publication types