Report
Teaser, summary, work performed and final results
Periodic Reporting for period 3 - MULTICONNECT (Imaging Brain Circuits to Decode Brain Computations: Multimodal Multiscale Imaging of Cortical Microcircuits to Model Predictive Coding in Human Vision)
Teaser
The human brain is one of the largest and most complex biological networks known to exist. It consists of a large number of interacting circuits, which play a crucial role determining the computations it can perform and, thus, in enabling our cognitive abilities. The...
Summary
The human brain is one of the largest and most complex biological networks known to exist. It consists of a large number of interacting circuits, which play a crucial role determining the computations it can perform and, thus, in enabling our cognitive abilities. The architecture of these circuits in humans, and therefore the computational basis of human cognition, remains largely unknown. The present proposal will focus on the question how human cortical microcircuits contribute to computation and cognition, and this question will be tackled from both a structural and a functional perspective.
The central idea of a canonical cortical microcircuit is that the neocortex replicates (with small variations) a basic columnar circuit design capable of performing computations. Recently, it has been suggested that the nature of these computations is to implement predictive coding principles , opening an important way to connect the neocortex’s circuits to the neocortex’s computations. Predictive coding theory posits that the brain is proactive and predictive at all stages of its processing and commits processing resources accordingly. In particular, the allocation of resources depends on the difference between input and expectation. Cortical areas continuously compute predictions which are sent back to hierarchically ‘lower’ cortical areas. For instance, in visual motion perception the human V5/MT+ area is thought to predict where the stimulus occurs next in the visual field, given its motion. By feedback projections, it sends this prediction to the relevant topographic location of V1 where it is compared to bottom-up visual input from the thalamus. The ideas central to predictive coding have even been formulated into an organizing principle of the entire brain . Therefore, the present proposal’s overall goal is to increase insight into the role of human cortical microcircuit properties in predictive coding.
A structural and functional assessment of microcircuitry in the human brain only recently came within the realm of possibilities, thanks to the development of magnetic resonance imaging (MRI) at ultra-high field-strengths (UHF) of 7T and above. Post-mortem structural imaging of human brain wiring with diffusion MRI has proven possible in small tissue samples, even at the spatial scale of hundreds of micrometers. This has brought mesocopic human connectomics in large tissue samples, possibly the whole human brain, into reach. In-vivo functional imaging of human cortical activity, spatially resolved for cortical columns and layers, was also recently shown to be feasible. When UHF structural and functional MR imaging are combined, this has the exciting potential of imaging the connections and activity of different components within the cortical microcircuit.
The central aim of this proposal is to image human cortical connectivity at multiple spatial scales in order to understand human cortical computations. The core hypothesis is that the variations in predictive coding computations performed by human cortical microcircuits in different visual areas are grounded in variations in their microcircuit connectivity. As a central case-study, this proposal investigates human visual apparent motion in the dorsal cortical visual system (dorsal stream), particularly in human areas V1/2/3 and V5/MT+. Achieving the aims would significantly advance our understanding of how cortical microcircuits compute. Moreover, the proposed project will deliver important new reference data for graph analytical characterizations of the human connectome and generative models of human cortical dynamics in the resting state. It will also inform (modelling) studies of task-based human cortical processing in health or after brain damage.
Work performed
The research program is organized in two workpackages, WP I and WP II. Milestones planned for realisation Mid-term were within WP I with the following aims and objectives:
Aim 1 (WP I): imaging the structural connections of human neocortical microcircuits
WP I progresses steadily down the spatial scales from the mesoscale to the microscale. The main method is ex-vivo UHF diffusion MRI. Entering the microscale is achieved by using advanced light microscopy techniques. Specific project objectives in the human dorsal stream are:
Project 1: Mesoscale imaging of white matter connectivity between areas
In this project the following Milestones were achieved Mid-term:
Milestone 1.1 (Human whole-brain mesoscale WM connectome (<500um res))
Milestone 1.2 (Human dorsal stream mesoscale WM connectome (<400um res))
Project 2: Mesoscale imaging of layered intracortical connectivity
In this project the following Milestones were achieved Mid-term:
Milestone 2.0 (Prep: Building, optimizing and testing tissue-slab MRI RF-coil) The planar tissue-slab coil was contructed and in the testing phase. As a crucial preliminary result, a cylindrical 9.4T RF-coil for post mortem human tissue scanning at 9.4T was developed with accompanying publication, and 60um isotropic human occipital lobe MRI was achieved.
Project 3: Microscale imaging of microcircuit connectivity statistics
Milestone 3.0 (Prep: optimizing tissue clearing and deep fluorescence imaging): This Milestone was realized ahead of time. An optical clearing and labelling protocol for cytoarchitecture characterization of human brain tissue samples was developed (MASH: Multiscale Architecture Staining of Human cortex)..
Additionally:
Two specific RF pulse sequences for post mortem human brain tissue diffusion MRI at 7T and 9.4T were developed: kT-dSSFP and kT-dSTEAM.
Two methods for in vivo analysis of human brain connectivity with MRI were developed within work in the Action and published.
Final results
The research program is organized in two workpackages, WP I and WP II. Milestones planned for realisation in the second half of the Action are mostly within WP II with the following aims and objectives:
Aim 2 (WP II): modeling how microcircuits support predictive coding computations
WP II progresses back up the spatial scales from the microscale to the mesoscale. The main method is in-vivo UHF functional MRI. Bridging structure to function is achieved with computational neural network modelling. Specific project objectives in the human dorsal stream are:
Project 4: Simulating predictive coding computations with cortical microcircuits
Project 5: Testing realistic predictive coding models with UHF human fMRI
Furthermore, a new variation of lightsheet microscope, the ct-dSPIM was conceptualized and developed which can image very large human brain tissue samples. Results from a fully functional ct-diSPIM are expected in the remainder (second half) of the Action.