Ultrafast fMRI Reveals Laminar Specificity of Hemodynamic Profiles in Primary Somatosensory Cortex

1348 

Abstract Submission 

Xiangnan Tian¹, Nadia Bouraoud², Stephen Dodd², Alan Koretsky², Zhiwei Ma¹

¹ShanghaiTech University, Shanghai, China, ²National Institutes of Health, Bethesda, MD, USA

Xiangnan Tian  
ShanghaiTech University
Shanghai, China

Nadia Bouraoud  
National Institutes of Health
Bethesda, MD, USA

Stephen Dodd  
National Institutes of Health
Bethesda, MD, USA

Alan Koretsky  
National Institutes of Health
Bethesda, MD, USA

Zhiwei Ma  
ShanghaiTech University
Shanghai, China

Recent advancements in ultra-high-field MRI have significantly improved fMRI, enabling detailed studies of individual cortical layers. This has provided insights into the timing of brain activity. Different layers of the cortex have unique connections and interactions between neurons and blood vessels. This complex relationship reflects how neuronal activity is linked to the BOLD signal, with each layer having its specific temporal profile. The fMRI activation initiates in mid-cortical layers and subsequently extends to other layers [1,2]. Despite the strides in studying laminar-specific fMRI onset, other laminar temporal features of hemodynamic profiles, such as fMRI signal offset, are yet to be understood. Here we utilized ultrafast fMRI to capture the temporal variations of the BOLD signal across different cortical depths in rat primary somatosensory cortex to reveal laminar-specific hemodynamic profiles.

Four alpha-chloralose anesthetized Sprague-Dawley rats were imaged using an 11.7 T MRI scanner to examine fMRI responses to forepaw stimulation. For each rat, a BOLD fMRI run with a TR of 1500 ms (voxel size = 0.17 mm×0.17 mm×0.4 mm, slice number = 14) was acquired to localize S1FP based on its activation map. This guided subsequent fMRI runs with a TR of 200 ms (voxel size = 0.17 mm×0.17 mm×1.2 mm, slice number = 1) positioned on the slice in the center of the S1FP. Each of these fMRI runs included 60 blocks and the stimulation was on for 2 s within each block. In the preprocessing, NORDIC denoising was applied [3]. The MATLAB function 'imregister' was used for motion correction. The cortical depth of each voxel in the S1FP was derived using Laplace's equation [4-7]. The voxels were grouped into 10 geometric layers based on their cortical depths. A high-pass filtering of 0.0179 Hz was performed. Outlier blocks were excluded using autoencoders. The BOLD percent change data of the remaining blocks were averaged and then fitted using the double gamma function. The response height, full-width-at-half-maximum (FWHM), time-to-peak, and time constant of the BOLD signal of each geometric layer were derived based on the fitted curve. Here, the time constant was the time difference between the stimulation on time and the time when the fitted curve reached 1/e of the maximum of the curve on the decay phase. In addition, instead of using the fitted curve, the BOLD onset time was calculated using the time difference between the stimulation on time and the time when the BOLD signal was firstly two standard deviations above the baseline.

Fig. 1a shows a representative result of a task fMRI activation map in S1FP and its corresponding layerification. It also shows BOLD percent change maps per geometric layers of an individual block and the average of 922 blocks after outlier exclusion. The BOLD percent change curves per geometric layers are shown in Fig. 1b. The hemodynamic profiles are shown in Fig. 2. We observed distinct laminar BOLD responses to rat forepaw stimulation. The response height decreased from superficial to deep layers (Fig. 2b). The third geometric layer of S1FP exhibited the shortest FWHM, time-to-peak, onset, and time constant (Fig. 2c-f).

This study has revealed a laminar-specific timing in the BOLD response to sensory stimuli. Consistent with previous findings [1,2], the middle layers had the shortest onset time. The cortex at its 30% depth also showed the shortest time constant of the BOLD decay phase. Ultrafast fMRI with laminar spatial resolution enhances our grasp of the neurovascular dynamics across cortical depths, offering important implications for interpreting fMRI data.

Activation (eg. BOLD task-fMRI) ¹

Methods Development ²

Cortical Layers

Data analysis

FUNCTIONAL MRI

Somatosensory

Other - Layer fMRI