Estimating laminar neuronal response using a hemodynamic model of depth-dependent BOLD signal
Wednesday, June 12, 2019: 11:30 AM - 11:42 AM
Auditorium Parco Della Musica
Room: Theatre Studio
Martin Havlicek1, Sriranga Kashyap1, Federico De Martino1, Kamil Uludag2,3
1Maastricht University, Maastricht, Netherlands, 2Center for Neuroscience Imaging Research, Sungkyunkwan University, Suwon, Korea, Republic of, 3Techna Institute & Koerner Scientist in MR Imaging, University Health Network, Toronto, Ontario, Canada
The BOLD signal comprises both neuronal and vascular sources of variability . Particularly in laminar fMRI, vascular changes in lower cortical depths affect blood oxygenation and volume in the upper depths via intra-cortical ascending veins (AV), making it difficult to infer the neuronal laminar profiles from the fMRI data. Recently , we have introduced a new hemodynamic model of the laminar BOLD signal based on mass balance principles that accounts for vascular biases and reliably distinguishes neuronal and vascular effects at the mesoscopic scale. This represents a clear departure from previously introduced laminar models, which are either phenomenological  or represent only the steady-state . We have previously demonstrated that the proposed model can reproduce typical laminar BOLD responses, including its dynamic features . Here, we demonstrate how this model can be inverted using a Bayesian estimation framework with physiological and experimental constraints.
Our model of laminar BOLD signal is driven by laminar changes in activity of excitatory-inhibitory (E-I) neuronal units due to experimental manipulation followed by neurovascular coupling (NVC) . The hemodynamic model  is characterized by its local (venous signal of the local microvasculature) and non-local (AV carrying deoxyhemoglobin (dHb) changes from the lower depths) compartments, see Fig.1A. Simulations using this model were performed, considering stimulation according to a classical working memory paradigm (see Fig.1A). We modeled the laminar effect in deep cortical depths of action trials (A) compared to non-action trials (NA) . Additionally, to emphasize the spatial variability between cortical depths, we formed two peaks of equal amplitude in the lower and upper depths. 15 simulation trials were created for each condition. The laminar BOLD model was adjusted for a gradient-echo sequence at 7T. BOLD responses generated from these spatiotemporal profiles were assigned to anatomically defined depths at super-resolution (0.125mm), then down-sampled to typical high-resolution (0.75mm) with added Gaussian noise (SNR=Amp(activation)/Std(noise)=0.75). For model inversion, we considered the time-varying BOLD activity in 150 voxels averaged across all trials. The forward model was constructed for six cortical depths and inverted using a Variational Bayesian framework .
Simulated BOLD responses differed significantly from the original neuronal responses (see Fig.1B and C), exhibiting a typical increase towards the cortical surface [8,9] and lacking the spatiotemporal specificity of the neuronal response. The forward model was informed about the experimental manipulation using driving and modulatory inputs to all neuronal units and hemodynamic variation in laminar BOLD responses due to AV (see Fig.2A) was accounted for. The fit of the predicted BOLD responses to the simulated laminar data was good (see Fig.2B). The estimated laminar neuronal responses at six cortical depths formed a representative approximation of the original laminar neuronal profiles (see Fig.2C). This shows that the effect of AV was successfully removed.
In this study we have shown that, with the temporal constraints given by the experimental manipulation (i.e. the different laminar profile induced by different conditions) and spatiotemporal constraints given by the hemodynamic model of laminar BOLD, model inversion and estimation of the underlying neuronal activity is feasible. This was achieved using simulations under realistic conditions, accounting for information loss and blurring due the resolution of fMRI data. In the future, the simple neuronal units could be replaced by the laminar neuronal model of the cortical microcircuit  and the model could be used to invert functional responses obtained in a variety of paradigms (e.g. classical oddballs or omission responses).
BOLD fMRI 1
Modeling and Analysis Methods:
Methods Development 2
Physiology, Metabolism and Neurotransmission :
Cerebral Metabolism and Hemodynamics
Design and Analysis
HIGH FIELD MR
Other - Laminar
1|2Indicates the priority used for review
My abstract is being submitted as a Software Demonstration.
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Was any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using author date format
 Uludag K. (2018), ‘Hemodynamic modeling of laminar resolution fMRI’, ISMRM Paris.
 Heinzle, J. (2016), ‘A hemodynamic model for layered BOLD signals’, NeuroImage, vol. 125, pp. 556-570.
 Markuerkiaga, I. (2016), ‘A cortical vascular model for examining the specificity of the laminar BOLD signal’, NeuroImage, vol. 132, pp. 491-498.
 Havlicek, M. (2015), ‘Physiologically informed dynamic causal modeling of fMRI data’, NeuroImage, vol. 122, pp. 355-372.
 Finn, E. (2018), ‘Layer-dependent activity in human prefrontal cortex during working memory’, pre-print at bioRχiv, pp. 1-17
 Friston, K. (2006), ‘Variational free energy and the Laplace approximation’, NeuroImage, vol. 34, pp. 220-234.
 Kashyap, S. (2017), ‘Impact of acquisition and analysis strategies on cortical depth-dependent fMRI’, NeuroImage, vol. 168, pp. 332-344.
 De Martino, F. (2013), ‘Cortical Depth Dependent Functional Responses in Humans at 7T: Improved Specificity with 3D GRASE’, PlosOne, vol. 8, pp. e60514.
 Pinotsis, D. (2017), ‘Linking canonical microcircuits and neuronal activity: Dynamic causal modelling of laminar recordings’, NeuroImage, vol. 146, pp. 355-366.