Analysis of sub-millimetre resolution fMRI for laminar dissection of the brain function
Project Overview
The project aims to advance the understanding of the human cerebral cortex by developing high-resolution fMRI sequences capable of analysing its intricate laminar structures. The cortex plays a crucial role in processing and generating signals that shape behaviour, and even at rest, certain areas exhibit synchronized activity measurable by fMRI. However, the involvement of the six cortical laminae in resting-state networks could not be studied due to the low resolution of standard fMRI sequences that only allows the cortex to be modelled as a two-dimensional structure, neglecting its depth.
Colleagues from INM-4 addressed these limitations by developing advanced high-resolution fMRI sequences (Yun et al., 2022) that maintain the field-of-view while providing depth-specific insights into resting-state networks (Pais-Roldan et al., 2023). This new methodology has shown promising results in characterizing functional connectivity with respect to cortical depths. However, pre-processing and analysing high-resolution whole-cortex fMRI data present several challenges, especially when using BOLD (blood oxygenation level dependent) contrast as required for this method.
As one of these challenges is the co-registration of the structural and functional sequence, a modified structural sequence is being tested to align structural and functional images, reducing data loss. The objectives include re-acquiring data with matched structural images for more reliable cortical mapping, enhancing pre-processing and analysis methods to validate and improve original findings, and performing novel analyses to explore dynamic connectivity and integrate simulations for deeper insights into laminar networks. A second objective is the development of tools to improve cortical mapping in existing legacy data, which includes over 30 patients with significant scientific value.
This research also aims to explore the application of these sequences to study brain disorders, with preliminary analysis revealing varying depth-to-depth connections within resting-state networks. These efforts are intended to enhance the understanding of the cerebral cortex and its implications for brain health.
Our contribution
Since parts of the imaging pipeline require high-performance computing (HPC) resources, we support INM-4 by deploying and optimizing the neuroimaging pipeline on our HPC systems at JSC. We are developing a deep learning-based co-registration method to align the structural and functional sequences, addressing the specific challenges posed by high-resolution fMRI. These challenges include independent distortions in both sequences present in the legacy data. Additionally, we are creating a reliability measure for these distortions to mask misaligned areas in the images.
Our collaboration partner
This project is being conducted in collaboration with the MR Group of INM-4.
References
Yun SD, Pais-Roldán P, Palomero-Gallagher N, Shah NJ. Mapping of whole-cerebrum resting-state networks using ultra-high resolution acquisition protocols. Hum Brain Mapp. 2022 Aug 1;43(11):3386-3403. doi: 10.1002/hbm.25855.
Pais-Roldan, P., Yun, S. D., Palomero-Gallagher, N. and Shah, N. J. (2023) 'Cortical depth dependent human fMRI of resting-state networks using EPIK', Front Neurosci, 17, pp. 1151544. doi: 10.3389/fnins.2023.1151544
SDL Neuroscience Contact
SDL Neuroscience Team
Machine Learning and Data Analytics for Neuroimaging