This talk covers the applications of multi-echo (ME)-fMRI in combination with independent components analysis (ICA), called ME-ICA (Kundu et al., 2011). The ME-ICA approach to fMRI acquisition and analysis greatly increases the fidelity of BOLD fMRI while reducing the burden of artifacts across fMRI applications. Thus, the target audience of this talk includes all users of fMRI. Examples include users of resting state fMRI, task-based fMRI, pharmaco-fMRI, clinical fMRI of patients with lesions, and preclinical fMRI. The evidence presented indicates that the ME-fMRI approach expands the range of experiments that is practicable using fMRI.
Target Audience
This talk covers the applications of multi-echo (ME)-fMRI in combination with independent components analysis (ICA), called ME-ICA (Kundu et al., 2011). The ME-ICA approach to fMRI acquisition and analysis greatly increases the fidelity of BOLD fMRI while reducing the burden of artifacts across fMRI applications. Thus, the target audience of this talk includes all users of fMRI. Examples include users of resting state fMRI, task-based fMRI, pharmaco-fMRI, clinical fMRI of patients with lesions, and preclinical fMRI.Outcome / Objectives
The objectives of this lecture include gaining an understanding of the methods and benefits of ME-fMRI acquisition and ME-ICA analysis in terms of signal-to-noise ratio, contrast-to-noise ratio, and sensitivity to functional activation and connectivity. This involves learning which MRI platforms support ME-fMRI, what are the pulse sequences that enable ME-fMRI, and what are key considerations while configuring an ME-fMRI pulse sequence. Another objective of this talk is conveying how to use the ME-ICA software to transform raw data into preprocessed and denoised time series, including the use of the meica.py software. We will also discuss what are the considerations to make for using ME-ICA processed time series in conjunction with standard fMRI analysis software such as SPM. The talk will convey quantitative performance metrics of ME-ICA analyses and comparisons to single-echo fMRI, such that users can have a working knowledge of costs and benefits of the ME and multi-band multi-echo (MBME)-fMRI approaches.Purpose
The purpose of the talk is to inform fMRI users of how ME-ICA can enable robust fMRI for task-based and resting-state fMRI; at the single-subject and group levels of study; across MRI field strengths (1.5 T, 3 T, 7 T); in healthy individuals and patients of advanced age or with brain lesions; in pharmaco-fMRI experiments, and in human and animal fMRI experiments. Significant benefits of ME-ICA in applications of fMRI to epilepsy and to neuropsychiatric domains involving the orbitofrontal cortex and subcortical areas are emphasized.Methods
The acquisition methods involve multi-echo EPI on Siemens and GE platforms at 3 T as well as 1.5 T and 7T MRI for human imaging. For animal imaging, multi-echo EPI is demonstrated for data from Bruker MRI systems at 7 T (primate) and 11.7 T (rodent) field strengths. Multi-band multi-echo fMRI is demonstrated for results at 3 T and 7T MRI. All ME data are analyzed with ME-ICA, an analysis toolbox that implements fMRI processing, ICA decomposition, component classification and de-noising, in most cases agnostic to imaging parameters or the particular anatomy of the subject (young and old; normative and lesional brains, etc.).Results
The results discussed pertain to signal quality measures, functional network characteristics, effect sizes and/or statistical power estimates from ME-fMRI and ME-ICA of data across MRI field strengths, in human patients and controls and preclinical models. The findings at 3 T from resting-state data of healthy individuals show that ME-ICA leads to up a 4-fold increase in signal-to-noise ratio and control of type I error in group-level comparisons of seed-based resting-state connectivity (Kundu et al., 2013). Results from task-based fMRI at 3 T involve activations of subcortical areas and cortical regions such as medial prefrontal cortex (from a mentalizing task; Lombardo et al., 2016) Task activation after ME-ICA denoising shows overall increases in group-level statistical power for detecting block-design activations, in some cases more than doubling power. For event-related designs, increased effect sizes are reported, including from data acquired with cardiac gating (Gonzalez-Castillo et al., 2016). A comparison is made between ME-ICA and other techniques such as FSL FIX and ICA-AROMA for resting-state connectivity at 1.5 T (DiPasquale et al., 2017). Results show that ME-ICA preprocessing yields better signal-to-noise ratio than all single-echo analysis methods, and equal or better such metrics than manually trained and/or bandpass filtered denoising of T2*-weighted combinations of multi-echo time series. Pharmaco-fMRI results from 3 T MRI include oxytocin-related changes in subcortical connectivity (Bethlehem et al., 2017). A novel experiment involving administration of sevoflurane anesthesia in healthy individuals including the elderly is shown. This study reveals a doubling of the number of functional components at peak anesthetic dose, indicating that anesthesia may be inducing functional dysconnectivity between brain regions. Recent resting-state MRI at 7 T results from lesional and non-lesional epilepsies and matched healthy controls will demonstrate cortical and subcortical functional connectivity at high spatial resolution and time series differences in epilepsy. Preclinical data from high-field rodent and primate MRI show that ME-ICA can reliably detect a wide array of functional networks under isoflurane anesthesia, including the default mode and subcortical networks (Kundu et al., 2014).Discussion
The objective of ME-ICA is to improve the fidelity of fMRI by removing non-BOLD signals, while making fMRI analysis more physically principled. As a result, fMRI analysis becomes more straightforward and generalizable across resting state and task conditions, across MRI field strengths, and is generalizable across anatomical and physiological (e.g. drug-related) variations. Statistical power and effect sizes of key effects are increased reliably, and false positive (type I) errors are reduced. However, the ME approach does have limitations including of spatial resolution, which will be discussed, particularly in the context of 7 T and multi-band fMRI.Conclusion
The evidence presented indicates that the ME-fMRI approach expands the range of experiments that is practicable using fMRI. These findings suggest a compelling future role of the multi-echo approach in subject-level and clinical fMRI.[1] Bethlehem, R.A.I., Lombardo, M.V., Lai, M.C., Auyeung, B., Crockford, S., Deakin, J., Soubramanian, S., Sule, A., Kundu, P., Voon, V. and Baron-Cohen, S., 2017. Oxytocin enhances intrinsic corticostriatal functional connectivity in women. Translational Psychiatry (in press).[2] Dipasquale, O. Sethi, A., Laganà M. M., Baglio, F., Baselli, G., Kundu, P., Harrison, N. A., Cercignani, M. (2017) Comparing resting state fMRI de-noising approaches using multi- and single-echo acquisitions. PLoS One (in press). [3] Gonzalez-Castillo, J., Panwar, P., Buchanan, L.C., Caballero-Gaudes, C., Handwerker, D.A., Jangraw, D.C., Zachariou, V., Inati, S., Roopchansingh, V., Derbyshire, J.A., et al. (2016). Evaluation of multi-echo ICA denoising for task based fMRI studies: Block designs, rapid event-related designs, and cardiac-gated fMRI. NeuroImage 141, 452–468. [4] Kundu, P., Inati, S.J., Evans, J.W., Luh, W.M., and Bandettini, P.A. (2011). Differentiating BOLD and non-BOLD signals in fMRI time series using multi-echo EPI. NeuroImage. [5] Kundu, P., Brenowitz, N.D., Voon, V., Worbe, Y., Vértes, P.E., Inati, S.J., Saad, Z.S., Bandettini, P.A., and Bullmore, E.T. (2013). Integrated strategy for improving functional connectivity mapping using multiecho fMRI. Proc. Natl. Acad. Sci. 201301725. [6] Kundu, P., Santin, M.D., Bandettini, P.A., Bullmore, E.T., and Petiet, A. (2014). Differentiating BOLD and non-BOLD signals in fMRI time series from anesthetized rats using multi-echo EPI at 11.7 T. NeuroImage 102 Pt 2, 861–874. [7] Lombardo, M.V., Auyeung, B., Holt, R.J., Waldman, J., Ruigrok, A.N.V., Mooney, N., Bullmore, E.T., Baron-Cohen, S., and Kundu, P. (2016). Improving effect size estimation and statistical power with multi-echo fMRI and its impact on understanding the neural systems supporting mentalizing. NeuroImage.