Validating Linear Systems Analysis for Laminar fMRI: Temporal Additivity for Stimulus Duration Manipulations

Jelle A Van Dijk,Alessio Fracasso,Serge O Dumoulin,Natalia Petridou

doi:10.1007/s10548-020-00808-y

Abstract

Advancements in ultra-high field (7 T and higher) magnetic resonance imaging (MRI) scanners have made it possible to investigate both the structure and function of the human brain at a sub-millimeter scale. As neuronal feedforward and feedback information arrives in different layers, sub-millimeter functional MRI has the potential to uncover information processing between cortical micro-circuits across cortical depth, i.e. laminar fMRI. For nearly all conventional fMRI analyses, the main assumption is that the relationship between local neuronal activity and the blood oxygenation level dependent (BOLD) signal adheres to the principles of linear systems theory. For laminar fMRI, however, directional blood pooling across cortical depth stemming from the anatomy of the cortical vasculature, potentially violates these linear system assumptions, thereby complicating analysis and interpretation. Here we assess whether the temporal additivity requirement of linear systems theory holds for laminar fMRI. We measured responses elicited by viewing stimuli presented for different durations and evaluated how well the responses to shorter durations predicted those elicited by longer durations. We find that BOLD response predictions are consistently good predictors for observed responses, across all cortical depths, and in all measured visual field maps (V1, V2, and V3). Our results suggest that the temporal additivity assumption for linear systems theory holds for laminar fMRI. We thus show that the temporal additivity assumption holds across cortical depth for sub-millimeter gradient-echo BOLD fMRI in early visual cortex.

Highlights

Magnetic resonance imaging (MRI) is the dominant method to non-invasively study the structure and function of the living human brain
While draining effects across cortical depth pose a challenge to laminar functional MRI (fMRI), these effects are largely irrelevant for fMRI at conventional resolutions (> > 1 mm3), as one voxel at these resolutions typically spans the thickness of the cortex
While we have previously shown that the linear scaling assumption across cortical depth holds for gradient echo (GRE) blood oxygenation level dependent (BOLD) in the visual cortex, the temporal additivity assumption remains to be evaluated

Summary

Introduction

Magnetic resonance imaging (MRI) is the dominant method to non-invasively study the structure and function of the living human brain. The fMRI signal at a specific cortical depth does consist of measurements of the hemodynamic consequences of local neuronal activity (Heeger et al 2000; Logothetis 2002), and of hemodynamic changes at underlying cortical depths, closer to the gray-white matter surface The weighting of these two factors—local neuronal activity in micro-vessels, and draining effects from micro- and macro-vasculature- is dependent upon the specific acquisition methods used (Kim and Ogawa 2012; Petridou and Siero 2017). While draining effects across cortical depth pose a challenge to laminar fMRI, these effects are largely irrelevant for fMRI at conventional resolutions (> > 1 mm3), as one voxel at these resolutions typically spans (most of) the thickness of the cortex

Methods

Results

Discussion

Conclusion