Single-Trial Characterization of BOLD fMRI Responses by Self-Organizing Neural Networks

Functional magnetic resonance imaging (fMRI) plays a dominant role in human brain mapping studies. As of today, no standards exist for processing fMRI data and analysis techniques are often associated to the different stimulation paradigms used to acquire functional data. Being able to estimate the hemodynamic response following a single execution of a task permits to characterize its relationship to different aspects of the stimulus, and of the subject’s performance. This works is aimed to test a strategy for the characterization of single trial-related BOLD fMRI responses based on the self-organizing maps method of Kohonen (SOMs). Analysis have been carried out on synthetic fMRI images modeling activation and on data from a single-event fMRI experiment on one human subject performing a basic motor task. Results were able to define the potentiality range for this data-driven methodology in monitoring the evolution of the BOLD response deriving from a single stimulation.