ICASENSE: Sensitivity mapping using Independent Component Analysis for parallel Magnetic Resonance Imaging

Parallel Magnetic Resonance Imaging (MRI) methods employ receiver coils sensitivities to reduce imaging time: reconstruction algorithms need RF field maps which must be measured or estimated. Assuming statistical independence of different regions in a MR image, we consider the sensitivity estimation as a Blind Source Separation (BSS) problem that can be solved with Independent Component Analysis (ICA). This new formulation permits sensitivity maps extraction from only one MR acquisition, without calibration step or acquisition of additional k-space lines. Simulation results are presented for sensitivity encoded (SENSE) MR images, proving that sensitivity data can be extracted from statistical properties of the image, using the method ICASENSE.