Perfusion quantification of the whole lung using singular value decomposition with optimized threshold

F. Risse, C. Fink, S. Ley, H-U. Kauczor, L. R. Schad Department of Medical Physics in Radiology, Deutsches Krebsforschungszentrum, Heidelberg, Germany, Department of Radiology, Deutsches Krebsforschungszentrum, Heidelberg, Germany Introduction Contrast-enhanced 3D MRI offers the capability to assess lung perfusion non-invasively and without radiation exposure [1]. Recently, parallel imaging techniques allow the acquisition of images of the entire lung in acceptable temporal resolution [2]. As a result it is now possible to track the first pass of an injected bolus of contrast agent in the whole lung. A deconvolution method is required for a model-independent evaluation of the lung perfusion based on the indicator dilution theory. Singular value decomposition (SVD) has shown its potential for the analysis of bolus tracking experiments in the brain [3]. Advantages of this method are the modelindependence and its ability to reduce the noise contribution due to a singular value threshold. Unlike brain perfusion measurements, there is a low signal-to-noise ratio (SNR) in the lung which also varies due to the relative tissue inhomogeneity. Thus a fixed threshold can lead to inaccurate values. It is therefore necessary to optimize the threshold for every deconvolution process considering the SNR. The aim of this work was to optimize the singular value threshold for the application of SVD in lung perfusion quantification and to evaluate of the method in 3D whole lung data sets.