1Radiological Physics, University Hospital of Basel, Basel, Switzerland, 2Department of Biomedical Engineering, University of Basel, Basel, Switzerland
In the contemporary Fourier decomposition lung MRI, time-resolved registered 2D image series are Fourier transformed to identify in a power spectrum the underlying respiratory and cardiac frequencies. Subsequently, the amplitudes corresponding to the respiratory and cardiac motion are extracted voxel-wise to eventually produce ventilation and perfusion images. However, the analysis of truncated oscillatory signals and the peak search in the Fourier spectrum is usually very unstable and inaccurate. Here, we propose to use a robust and fully-automated method of signal analysis using a matrix pencil decomposition in combination with a linear least squares analysis for improved quantitative pulmonary function assessment.