Structural Equation Modelling of Inversion-Recovery-BOLD laminar fMRI

Poster No:


Submission Type:

Abstract Submission 


Jiewon Kang1, Ido Tavor2, Yaniv Assaf2, Mark Woolrich1, Saad Jbabdi1


1Wellcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom, 2Tel Aviv University, Tel Aviv, Israel

First Author:

Jiewon Kang  
Wellcome Centre for Integrative Neuroimaging, University of Oxford
Oxford, United Kingdom


Ido Tavor  
Tel Aviv University
Tel Aviv, Israel
Yaniv Assaf  
Tel Aviv University
Tel Aviv, Israel
Mark Woolrich  
Wellcome Centre for Integrative Neuroimaging, University of Oxford
Oxford, United Kingdom
Saad Jbabdi, PhD  
Wellcome Centre for Integrative Neuroimaging, University of Oxford
Oxford, United Kingdom


Laminar FMRI is an emerging sub-field in neuroimaging where attempts are being made at resolving brain activity in different cortical layers (Koopmans et al., 2011). Current approaches use high-field (e.g. 7T) scanners to boost SNR in high resolution (sub-millimetre) acquisitions. However, this comes at the cost of reduced brain coverage in addition to issues of transmit field inhomogeneities at high field. We have recently proposed an alternative approach in which instead of resolving cortical layers spatially, we modulate the signal in a layer-dependent manner using inversion pulses prior to each BOLD acquisition (Tavor et al., 2017). This allowed us to acquire whole-brain, laminar-modulated FMRI.

Here we propose an effective connectivity model to infer laminar connectivity between different brain areas. We use an extension of structural equation models (SEM), whereby the structural equations pertain to unobserved laminar activity, to which we add a partial volume (observation) model to capture the combined effects of how different cortical layers respond to the inversion times. We show in simulations that the model is invertible, and we apply it to resting-state inversion-modulated spin echo BOLD acquisitions (at 3T), where we test for known feedforward vs feedback hierarchical organisation in the visual cortex. We show that our model recovers the correct known cortical hierarchy and outperforms competing models with equal complexity but alternative hierarchies.


Figure 1a-c illustrates the SEM and observation (partial volume) model. Partial volumes were calculated on the basis of a distribution of T1 relaxations within the cortex of [850, 1000, 1200]ms for [lower, middle, upper] layers based on data from (Barazany & Assaf, 2012). Simulations were conducted for pairwise connectivity between regions containing 2-3 layers.
Resting-state data were acquired for five healthy subjects in a 3T Siemens MAGNETOM Prisma system with inversion-recovery spin-echo echo planar imaging. Four inversion times (300, 480, 690, 810)ms were selected to modulate signals in different layers. Other scan parameters were TR/TE = 3000/45ms, in-plane resolution 3.1x3.1mm, slice thickness 4.5mm. After motion-correction, de-drifting, low-pass filtering (.1Hz), spatial smoothing, motion regression, and non-linear registration to MNI space in FSL; we chose 3 Brodmann areas (BA17,BA18,BA19) for analysis with our SEM approach.

Figure 1d shows the structural model which uses canonical feedforward and feedback connections in a 3-layer model (Felleman & Van Essen, 1991). We apply the same model to the data but permute the order of the 3 regions as means of comparing hierarchical models of equal complexity (Fig 1e). Model fitting was done with Metropolis Hastings sampling under the standard SEM likelihood, and uniform priors on the connection weights and noise variance.
Supporting Image: fig1_ohbm.png


Figure 2a,b show the results of the simulations, demonstrating that the model is identifiable with this combination of inversion times and partial volumes. The correct model also produces the lowest error for the weight parameters (Fig 2b). In real data, we show that the correct hierarchical model (BA17 lower than BA18 lower than BA19) has the best fit in all five subjects compared to alternative models of equal complexity (i.e. number of free parameters) with shuffled regional ordering (Fig 2c).
Supporting Image: ohbm_fig2.png


We have shown that we can use structural equation modelling to analyse laminar connectivity in an inversion-modulated BOLD acquisition. This opens the potential to modelling whole brain laminar connectivity and discovering cortical hierarchies in the human brain in the future.

Modeling and Analysis Methods:

fMRI Connectivity and Network Modeling 1
Methods Development
Task-Independent and Resting-State Analysis

Neuroanatomy, Physiology, Metabolism and Neurotransmission:

Cortical Anatomy and Brain Mapping 2

Novel Imaging Acquisition Methods:



Cortical Layers

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.

Resting state

Healthy subjects only or patients (note that patient studies may also involve healthy subjects):

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.

Not applicable

Please indicate which methods were used in your research:

Functional MRI

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

Barazany, D. (2012), ‘Visualization of cortical lamination patterns with magnetic resonance imaging’, Cerebral Cortex, vol. 22, pp. 2016–2023.
Felleman, D. J. (1991), ‘Distributed hierarchical processing in the primate cerebral cortex’, Cerebral Cortex, vol. 1, pp. 1-47.
Koopmans, P. J. (2011), ‘Multi-echo fMRI of the cortical laminae in humans at 7T', NeuroImage, vol. 56, pp. 1276-1285.
Tavor, I. (2017, June 25-29), ‘Whole-Brain Laminar Functional Connectivity with Inversion-Recovery FMRI [abstract]', Annual Meeting of the Organization for Human Brain Mapping