Characterization and compensation of artifacts in functional magnetic resonance imaging

Functional magnetic resonance imaging (fMRI) is a non-invasive neuroimaging modality used to study brain function by acquiring a volumetric time series. Various types of noise such as system noise and physiological noise are significant confounds in fMRI experiments that must be characterized and corrected to improve experimental results. A source of noise in variable repetition time fMRI is gradient coil heating and eddy currents. Typically, fMRI experiments use a constant repetition time (TR) for the acquisition of a volumetric time series, but variable TR experiments have been employed that enables researchers to perform studies not feasible with a constant TR. With a variable TR paradigm, volumes are collected at non-uniform time intervals, causing the temperature state of the gradient coils and eddy currents to vary as a function of TR. The resulting signal fluctuations represent an observable and deleterious confound for variable TR fMRI. This confound was characterized in this work using a novel radio-frequency disabled gradient sequence paradigm, and a modified projection method was developed to effectively correct for these adverse signal fluctuations. Another source of artifactual signal fluctuations in fMRI data is acoustic imaging noise produced from audible vibrations of gradient coils during image acquisition. This acoustic imaging noise not only interferes with the subject’s perception of desired stimuli in auditory fMRI, but also induces hemodynamic responses that interfere with the desired responses. In order to develop a compensation procedure for acoustic imaging noise, the hemodynamic response arising from pure acoustic imaging noise must first be characterized as a function of noise duration as well as noise intensity. Results show that the human auditory cortex exhibited increased response amplitude as well as response spatial extent to acoustic imaging noise of longer duration and intensity. Responses were strongly nonlinear for brief duration acoustic imaging noise, but became approximately linear with longer duration acoustic imaging noise. Improved characterization of hemodynamic responses to acoustic imaging noise and assessment of response linearity conducted in this work will enhance the ability to model responses under a variety of acoustic conditions in auditory fMRI, potentially enabling the development of a model to account for losses in sensitivity arising from the adverse interaction of acoustic imaging noise-induced responses with desired stimulus-induced responses