Differentiating BOLD and non-BOLD signals in fMRI time series using multi-echo EPI - PubMed (original) (raw)

Differentiating BOLD and non-BOLD signals in fMRI time series using multi-echo EPI

Prantik Kundu et al. Neuroimage. 2012.

Abstract

A central challenge in the fMRI based study of functional connectivity is distinguishing neuronally related signal fluctuations from the effects of motion, physiology, and other nuisance sources. Conventional techniques for removing nuisance effects include modeling of noise time courses based on external measurements followed by temporal filtering. These techniques have limited effectiveness. Previous studies have shown using multi-echo fMRI that neuronally related fluctuations are Blood Oxygen Level Dependent (BOLD) signals that can be characterized in terms of changes in R(2)* and initial signal intensity (S(0)) based on the analysis of echo-time (TE) dependence. We hypothesized that if TE-dependence could be used to differentiate BOLD and non-BOLD signals, non-BOLD signal could be removed to denoise data without conventional noise modeling. To test this hypothesis, whole brain multi-echo data were acquired at 3 TEs and decomposed with Independent Components Analysis (ICA) after spatially concatenating data across space and TE. Components were analyzed for the degree to which their signal changes fit models for R(2)* and S(0) change, and summary scores were developed to characterize each component as BOLD-like or not BOLD-like. These scores clearly differentiated BOLD-like "functional network" components from non BOLD-like components related to motion, pulsatility, and other nuisance effects. Using non BOLD-like component time courses as noise regressors dramatically improved seed-based correlation mapping by reducing the effects of high and low frequency non-BOLD fluctuations. A comparison with seed-based correlation mapping using conventional noise regressors demonstrated the superiority of the proposed technique for both individual and group level seed-based connectivity analysis, especially in mapping subcortical-cortical connectivity. The differentiation of BOLD and non-BOLD components based on TE-dependence was highly robust, which allowed for the identification of BOLD-like components and the removal of non BOLD-like components to be implemented as a fully automated procedure.

Published by Elsevier Inc.

PubMed Disclaimer

Figures

Figure 1

Figure 1

Shown are three echo simulations of BOLD (R2* change) and non-BOLD (S0 change) signals as a function of echo time (TE). The left column shows how the signal evolves for non-BOLD effects and the right column shows how the signal evolves for BOLD effects. The top row shows the signal during state x (no activation) and state y (activation). This top row demonstrates how the decay curves between rest and activation change in a different manner depending on if there is a change in (a) S0 or (b) R2*. The middle row shows the difference (y − x) signal for (c) change in S0, and (d) change in R2*. The bottom row shows the percent signal change (y−x)/0.5(x+y) for (e) change in S0, and (f) change in R2*.

Figure 2

Figure 2

(a) Multi-echo EPI images acquired at TE values of 15ms, 39ms, and 63ms. Image intensity decreases exponentially with TE. (b). Left: multi-echo EPI time courses from a voxel in visual cortex (center of yellow box in (a)) during periodic visual stimulation plotted as percent signal change. Right: percent signal change amplitude as a function of TE (black), with linear fit, i.e. change in R2* (gold). The fit is significant (p<0.01), with ΔT2*=0.3ms. (c) Left: multi-echo EPI time courses from a single precuneus voxel (center of white box in (a)) during rest, plotted as percent signal change. TE is indicated by the color. Right: percent signal change amplitude as a function of TE (black), with linear fit, i.e. change in T2* (gold). The fit is significant (p<0.01), with dT2*=0.3ms.

Figure 3

Figure 3

TE-dependence maps of ICA components from ME-ICA. (a) A BOLD-like and (b) a non BOLD-like component. For each component (a and b), the left panel shows percent signal change maps for three TEs 15ms, 39ms, 63ms (above), and the component time course (below). The right panel shows results of fitting to the ΔT2* change model (above) and the S0 change model (below). Goodness of fit maps, F{ΔT2*} and F{ΔS0}, are used to threshold parameter maps α<0.01 (p<0.05). (a) BOLD-like component: High percent signal change in gray matter scales linearly with TE. The component time course exhibits low frequency fluctuations. (b) non BOLD-like component: High percent signal change at the edge of brain is constant with TE. The component time course exhibits high frequency fluctuations.

Figure 4

Figure 4

For a representative subject, κ score vs (a) ICA rank (variance explained), and (b) rank by κ (κ spectrum). The κ spectrum, is an L-curve with two distinct regimes: high κ (κ >20) and low κ (κ <20), with low κ components on a linear tail. (c) κ spectra for 8 subjects. (d) First 12 ME-ICA components ranked by κ for a representative subject. Each panel shows the time course and thresholded ΔR2* map. Components are annotated with κ-score, ρ-score, and ICA component number. All high κ components are clearly functional networks.

Figure 5

Figure 5

For a representative subject, ρ score vs (a) ICA rank (variance explained), and (b) ρ rank (ρ spectrum). The ρ spectrum, like the k-spectrum, is an L-curve with two distinct regimes: high ρ (appx. ρ >20) and a linear tail with low ρ (appx. ρ <20). (c) ρ spectra for 8 subjects. (d) First 8 ME-ICA components ranked by ρ for a representative subject. Each panel shows the time course and thresholded %ΔS0 map. Components are annotated with κ-score, ρscore, and ICA component number. All high ρ components are clearly artifacts.

Figure 6

Figure 6

Components with κ scores near κ thresholds are correlated to low-frequency RVT time courses. Components are annotated with κ score, ρ score, and ICA component number. TE-dependence maps for ΔR2* and ΔS0 models show high ΔR2* localized to non-gray matter regions.

Figure 7

Figure 7

Signal from three regions of interest from a representative subject: the right insula, left hippocampus, and brainstem. (a) shows de-noising by removing: drifts only; drifts, physiology, and motion (the standard); drifts and low κ components (ME-ICA). (b) Seed based connectivity measured by R2 and T values, with baseline regression for: motion and physiology, then band pass filtering for 0.02–0.1 Hz; drifts and low κ components, without band pass filtering.

Figure 8

Figure 8

Group T-maps of subcortical connectivity with hippocampus and brainstem computed from individual maps with baseline regression for: motion and physiology, then band pass filtering; drifts and low κ components, without band pass filtering.

Figure 9

Figure 9

Group maps for subcortical connectivity with hippocampus and brainstem after removal of drifts and low κ and/or high ρ components. Overlay is map of mean Z-value, thresholded by T-value corresponding to FDR corrected p<10−4. Underlay is template brain in Talairach space.

Similar articles

Cited by

References

    1. Bandettini PA, Wong EC, Hinks RS, Tikofsky RS, Hyde JS. Time course EPI of human brain function during task activation. Magnetic Resonance in Medicine. 1992;25:390–397. - PubMed
    1. Bandettini PA, Wong EC, Jesmanowicz A, Hinkst RS, Hyde JS. Spin-Echo and Graciient-E Brain Activation using BOLD a Comparative Study at 1.5 T. NMR in Biomedicine. 1994;7:12–20. - PubMed
    1. Barth M, Reichenbach J, Venkatesan R, Moser E, Haacke E. High-resolution, multiple gradient-echo functional MRI at 1.5 T. Magnetic resonance imaging. 1999;17:321–329. - PubMed
    1. Barth M, Windischberger C, Klarhöfer M, Moser E. Characterization of BOLD activation in multi-echo fMRI data using fuzzy cluster analysis and a comparison with quantitative modeling. NMR in Biomedicine. 2001;14:484–489. - PubMed
    1. Beckmann C, Smith S. Probabilistic independent component analysis for functional magnetic resonance imaging. Medical Imaging, IEEE Transactions on. 2004a;23:137–152. - PubMed

Publication types

MeSH terms

LinkOut - more resources