Resting state network estimation in individual subjects.

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Neuroimaging

Background:

• Resting state functional magnetic resonance imaging (fMRI) is crucial for studying brain networks in health and disease.
• Accurate mapping of resting state networks (RSNs) in individuals is essential for understanding cognitive functions.
• Current methods for RSN topography may lack precision at the individual level.

Purpose of the Study:

• To develop a reliable method for computing individual resting state network (RSN) topography.
• To train a supervised classifier to associate functional connectivity maps with specific RSN identities.
• To evaluate the performance of the developed method against existing techniques.

Main Methods:

• Trained a multi-layer perceptron (MLP) classifier using blood oxygen level dependent (BOLD) correlation maps from predefined seeds.
• Classified RSNs based on correlation maps and propagated these through the MLP for whole-brain topography estimation.
• Compared MLP performance against dual regression and linear discriminant analysis for RSN mapping.

Main Results:

• Hard classification of RSNs using a priori seeds demonstrated high reliability across participants.
• Continuous estimates of RSN membership showed residual error, suggesting hierarchical network organization.
• The MLP accurately estimated RSN topography in individuals, including novel brain regions, outperforming alternative methods in spatial specificity.

Conclusions:

• The MLP-based approach provides a reliable and spatially specific method for individual RSN topography.
• This technique advances the ability to study the relationship between RSN organization and cognitive function or dysfunction.
• The findings support the hierarchical organization of RSNs and offer a superior tool for neuroimaging research.