CCASENSE: Canonical Correlation Analysis for Estimation of Sensitivity Maps for Fast MRI

Abstract: Magnetic Resonance Imaging is an established technology for both imaging and functional studies in clinical and research environments. The field is still very research intense. Two major research areas are acquisition time and signal quality. The last decade has provided tools for more efficient possibilities of trading these factors against each other through parallel imaging. In this thesis one parallel imaging method, Sensitivity Encoding for fast MRI (SENSE) is examined. An alternative solution CCASENSE is developed. CCASENSE reduces the acquisition time by estimating the sensitivity maps required for SENSE to work instead of running a reference scan. The estimation process is done by Blind Source Separation through Canonical Correlation Analysis. It is shown that CCASENSE appears to estimate the sensitivity maps better than ICASENSE which is a similar algorithm.