Improved acquisition and analysis methods for BOLD fMRI

Functional Magnetic Resonance Imaging (fMRI) is a noninvasive technique that acquires image sequences of the brain that are then analyzed to detect neuronal activity.;fMRI acquisitions are rapidly performed with a multislice single-shot scan trajectory because of its sensitivity to BOLD field variation and its high temporal resolution. Unfortunately, long readout duration and echo time in the single-shot trajectory cause distortion and susceptibility artifacts.;The first part of this thesis presents two new acquisition techniques: Spiral-In/Out and Interleaved Spiral-In/Out that reduce susceptibility-induced signal loss at air-tissue interface; they enable activation detection in ventral medial prefrontal cortex and amygdala where other common techniques falter.;Next we discuss parallel acquisition techniques (multiple receive-coils) that provide reduced acquisition time, hence reduction of susceptibility artifacts. A robust parallel acquisition technique, called sensitivity encoding (SENSE), reconstructs images using smoothed receive-coil sensitivity profiles. We demonstrate how retention of sensitivity profile noise (no smoothing) actually reduces thermal noise in reconstructed images. Neuronal activation at high spatial resolution can therefore be detected by our approach.;The most widely used fMRI technique, called BOLD (blood oxygenation level dependent) contrast, is based on changes in blood oxygenation consequent to changes in neuronal activity. These changes induce magnetic field variations in localized brain regions that modulate T2* relaxation time-constant.;Finally, we demonstrate a simple, fast, and flexible 1-norm minimization method to deconvolve haemodynamic response function (HRF) from response to fMRI task (not calibration) stimulus. Due to its variability, HRF of each individual person is required. An extra (time consuming, task limiting) calibration scan has commonly been performed before functional scans to capture an individual's HRF. This method, based on HRF being sparse in a wavelet domain, instead deconvolves HRF from fMRI task data by convex optimization.