Hemodynamic response measured with blood oxygenation level-dependent (BOLD) fMRI typically exhibit transients in the form of early-overshoot and post-stimulus undershoot. These transients originate from dynamic relationships between different physiological variables. They can be related to (1) active neuronal and metabolic processes reflecting changes in excitatory-inhibitory (E-I) balance; or (2) passive vascular venous blood volume changes due to vessel viscoelasticity. In this lecture, I will explain how dynamic physiological models, accounting for both active and passive mechanisms underlying BOLD response (BR) transients, can help us to study dynamic changes in E-I balance using fMRI data.
Target audience
The attendees will:
Content overview
Hemodynamic responses, in general, and the blood oxygenation level-dependent (BOLD) fMRI signal, in particular, provide an indirect measure of neuronal activity. There is a strong evidence that the BOLD response (BR) correlates well with post-synaptic changes, induced by changes in the excitatory and inhibitory (E-I) balance between active neuronal populations [1]. Changes in neuronal activity induce metabolic and vascular processes, resulting in hemodynamic response function (HRF) characteristic by temporal delay and spatiotemporal blurring with respect to neuronal activity. Typical BRs exhibit transients, such as the early-overshoot and post-stimulus undershoot (PSU) [2]. Many studies suggest that these dynamic features can be linked to transients in neuronal activity due to changes in E-I balance [3,4]. For example, a net increase in inhibitory activity after stimulus cessation can lead to decrease in cerebral blood flow (CBF) [5]. Other studies show that they can also result from dynamic uncoupling between CBF and oxygen metabolism (CMRO2) [6,7] or CBF and venous cerebral blood volume (vCBV) [8-10]. Further, it is possible that CBF and CMRO2 are driven in parallel by different aspects of neural activity [11]. This can result in a dynamic mismatch between CBF and CMRO2 responses and consequently create BR transients. We refer to neuronal and metabolic sources of BR transients as active mechanisms. On the other hand, especially for longer stimulus duration, the vCBV often lags behind CBF response due to vessel viscoelasticity, which also creates BR transients. We refer to this vascular source as passive mechanism. Practically speaking, several physiological mechanisms underlying BR transients can occur together. As a result, temporal features of the BOLD signal, such as response adaptation during stimulation or PSU, cannot be simply taken as a direct evidence of dynamic changes in E-I balance.
Nevertheless, physiologically informed dynamic causal models that take into account the above-mentioned dynamics of the physiological mechanisms underlying the BR [2,12] are designed to distinguish between active and passive mechanisms and can shed more light on underlying neuronal activity and E-I balance during response transients. This can be well demonstrated with arterial spin labeling data (ASL), which provide measurements of both CBF and BRs. If we experimentally modulate neuronal activity (e.g. using different type of visual stimuli), we should be able to observe differences in CBF response (including its transients) during two conditions. It is often the case that BR and its transients then also show difference between two conditions. Discrepancy in the size of response transients between CBF and BOLD, such as larger response adaption and/or PSU in the BR, can be attributed to CBF-CMRO2 or CBF-vCBV uncoupling. At this point, since changes in both CMRO2 and vCBV have very similar effect on BR, it might appear that they cannot be simply separated from each other. However, CBF-CMRO2 uncoupling relates to active mechanisms, which must to some extend reflect experimental modulation and changes in E-I balance. Thus, if the CBF-CMRO2 uncoupling contributes to BR transients, discrepancy in the size of transients between CBF and BR (formulated in terms of dynamic relationships), should be condition specific. On the other hand, if the passive CBF-vCBV uncoupling is the additional source of BR transients, then this should be condition (i.e. due to differences in neuronal activity) independent. Note that vCBV is still in dynamic (but passive) relationship with CBF, which is known to depend on stimulus duration and can differ between stimulation and post-stimulation periods. Therefore, using physiologically motivated models that can distinguish between active (neuronal) and passive (vascular) source of BR transients together with experimental manipulation can provide more accurate estimate of changes in neuronal activity from fMRI data. This is especially the case if both CBF and BOLD data are measured simultaneously [13]. In addition, also relying on only the BR but with clear experimental manipulations allows (based on logic above) separation of active and passive mechanisms [12,14].
In general, considering both of active and passive mechanisms underlying BR transients is an imperative to jointly model multi-contrast fMRI data, such as CBF&BOLD, CBV&BOLD or multi-echo BOLD approaches [13-15]. It can also play an important role in estimating the effective connectivity, leading to more reliable identification of connection strengths between brain regions [14].
[1] Logothetis, N.K., 2008. What we can do and what we cannot do with fMRI. Nature 453, 869–878.
[2] Buxton, R.B., Uludag, K., Dubowitz, D.J., Liu, T.T., 2004. Modeling the hemodynamic response to brain activation. NeuroImage 23, 220–S233.
[3] Shmuel, A., Augath, M., Oeltermann, A., Logothetis, N.K., 2006. Negative functional MRI response correlates with decreases in neuronal activity in monkey visual area V1. Nat. Neurosci. 9, 569–577.
[4] Sadaghiani, S., Ugurbil, K., Uludag, K., 2009. Neural activity-induced modulation of BOLD post-stimulus undershoot independent of the positive signal. Magn. Reson. Imaging 27, 1030–1038.
[5] Mullinger, K. J., Cherukara, M. T., Buxton, R. B., Francis, S. T., and Mayhew, S. D., 2017. Post-stimulus fMRI and EEG responses: Evidence for a neuronal origin hypothesised to be inhibitory. Neuroimage 157, 388–399.
[6] Frahm, J., Baudewig, J., Kallenberg, K., Kastrup, A., Merboldt, K.D., Dechent, P., 2008. The post-stimulation undershoot in BOLD fMRI of human brain is not caused by elevated cerebral blood volume. NeuroImage 40, 473–481.
[7] van Zijl, P.C.M., Hua, J., Lu, H., 2012. The BOLD post-stimulus undershoot, one of the most debated issues in fMRI. NeuroImage 62, 1092–1102.
[8] Mandeville, J.B., Marota, J.J.A., Ayata, C., Zaharchuk, G., Moskowitz, M.A., Rosen, B.R., Weisskoff, R.M., 1999. Evidence of a cerebrovascular postarteriole windkessel with delayed compliance. J. Cereb. Blood Flow Metabol. 19, 679–689.
[9] Chen, J.J., Pike, G.B., 2009. Origins of the BOLD post-stimulus undershoot. NeuroImage 46, 559–568.
[10] Buxton, R.B., Wong, E.C., Frank, L.R., 1998. Dynamics of blood flow and oxygenation changes during brain activation: the balloon model. Magnetic Reson. Med. 39, 855–864.
[11] Buxton, R.B., 2012. Dynamic models of BOLD contrast. NeuroImage 62, 953–961.
[12] Havlicek, M., Roebroeck, A., Friston, K., Gardumi, A., Ivanov, D., and Uludag, K., 2015. Physiologically informed dynamic causal modeling of fMRI data. Neuroimage 122, 355–372.
[13] Havlicek, M., Ivanov, D., Roebroeck, A., Uludag, K., 2017. Determining excitatory and inhibitory neuronal activity from multimodal fMRI data using a generative Hemodynamic Model. Front. Neurosci. 11, 1-20.
[14] Havlicek, M., Roebroeck, A., Friston, K. J., Gardumi, A., Ivanov, D., and Uludag, K., 2017. On the importance of modeling fMRI transients when estimating e ff ective connectivity: A dynamic causal modeling study using ASL data. NeuroImage 155, 217–233.
[15] Havlicek, M., Ivanov, D., Poser, B. A., and Uludag, K., 2017. Echo-time dependence of the BOLD response transients–a window into brain functional physiology. Neuroimage 159, 355–370.