Feedback and feedforward laminar modulation of human motion perception: a 7T fMRI study

2554 

Abstract Submission 

Alessandra Pizzuti¹, Omer Faruk Gulban^1,2, Laurentius (Renzo) Huber³, Judith Peters¹, Rainer Goebel¹

¹Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht, Netherland, ²Brain Innovation, Maastricht, Netherlands, ³Functional MRI Core Facility, National Institute of Mental Health, Bethesda, MD

Alessandra Pizzuti  
Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience
Maastricht, Netherland

Omer Faruk Gulban, Ph.D.  
Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience|Brain Innovation
Maastricht, Netherland|Maastricht, Netherlands

Laurentius (Renzo) Huber, PhD  
Functional MRI Core Facility, National Institute of Mental Health
Bethesda, MD

Judith Peters  
Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience
Maastricht, Netherland

Rainer Goebel  
Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience
Maastricht, Netherland

Ultra-high field fMRI (≥7 Tesla) enables sub-millimeter investigation of human conscious perception. Prior studies dissociated perceptual interpretation of visual motion in the motion sensitive area (hMT+) [1]. Yet, how perceived motion is constructed and modulated in the visual stream is unknown. Our 7T fMRI study explores how motion perception is orchestrated in visual areas (V1, V2, V3 and hMT+) and their laminar connectivity profiles. Using a bistable stimulus (physical and ambiguous motion quartet) [1], we dissociate feedforward from feedback signal integration and their neural correlates. Our results from V1 and hMT+ show differential laminar profiles for the two stimulus conditions.

The ambiguous quartet induced perception of horizontal or vertical movement through blinking squares at diagonally opposite corners (80s), alternating with a flicker condition (16s). Participants indicated percepts via a MR-compatible button box. In the physical quartet, squares moved along the same horizontal or vertical paths as targeted during the ambiguous motion. Five healthy participants underwent two 2-hour sessions using Siemens MAGNETOM Plus 7T at Scannexus (Maastricht, Netherlands). Per participant, we collected an MP2RAGE [4] (0.7 iso mm), a hMT+ functional localizer, 2 retinotopic mapping runs and 12 runs (12 min each) for the motion quartet stimulus (6 physical) using 2D gradient-echo (GE)-BOLD (0.8 iso mm, TR=2s) [5]. BOLD fMRI data (bilateral coverage) underwent preprocessing: slice time correction [6], motion correction [6], distortion correction [7], high-pass filter (5 cycles) [6] and co-registered to the anatomical images via boundary-based registration [6]. Two subjects were excluded due to motion. Anatomical images were bias field corrected [8], skull-stripped [7], upsampled to 0.35 iso mm. White (WM) and gray matter (GM) segmentation was performed [6] and manually corrected in ITK-SNAP [8]. GM cortical layers were defined by the equi-volume principle [9]. WM surfaces were reconstructed [6] and used to define the regions of interest (ROI). V1, V2, V3 ROIs were delineated using polar angle and eccentricity maps [10]. hMT+ ROI was defined using a separate visual motion localizer. A multi-run general linear model [6] characterized functional responses to physical and ambiguous conditions. Horizontal and vertical clusters (top 10% t-values, from the contrast: horizontal > vertical) in V1 and hMT+ were propagated to cover the whole cortical depth [9] and visualized on flattened GM [9]. Only voxels maintaining preference (horizontal or vertical) between two conditions (congruent voxels) were considered for laminar analysis (t-maps are upsampled to 0.35 iso mm).

Although our ambiguous stimulus elicits a robust motion percept, fMRI responses are consistently lower compared to the physical condition, suggesting reduced brain response without sensory input. In the physical condition, V1 shows a stronger response than hMT+ (Fig.1 B, F), aligning with its expected retinotopic organization. Conversely, fMRI modulation is similar across ROIs in the ambiguous condition (Fig.1-C, G). hMT+ exhibits a higher percentage of congruent voxels between conditions (64%) than V1 (46%) on average. Cortical depth analysis reveals a consistent V1 middle layer peak in the physical condition, hinting at deep layer modulation in the ambiguous condition (Fig.1-C,D and Fig2-A). Laminar profiles in hMT+ lack a clear pattern across participants (Fig.2), we observe a demodulation of the middle layer in only one participant (Fig.1-L).

In conclusion, our high-resolution BOLD fMRI study disentangles feedforward and feedback modulation in V1. Preliminary results confirm feedforward influence on the middle layer and feedback on the deep layers during motion perception, aligning with previous V1 studies [3]. Further analysis is crucial to address the draining vein effect, conduct statistical inference on layer profiles, and explore laminar behavior in V2 and V3.

Connectivity (eg. functional, effective, structural)

BOLD fMRI ²

Perception: Visual ¹

Consciousness

Cortical Columns

Cortical Layers

FUNCTIONAL MRI

HIGH FIELD MR

MRI

Perception

Vision