Cerebral blood flow measurement by dynamic contrast MRI using singular value decomposition with an adaptive threshold

Singular value decomposition (SVD) is a promising deconvolution technique for use in dynamic contrast agent magnetic resonance perfusion imaging. Computer simulations, however, show that the selection of the threshold for SVD affects the accuracy of the cerebral blood flow measurements and may distort the shape of the vascular residue function. In this report, a pixel-by-pixel thresholding method is proposed based on the signal-to-noise ratio of the concentration time curve at maximum concentration (SNRC). Monte Carlo simulations were used to determine the optimal threshold for different SNRC. This technique was used to analyze data from six healthy volunteers, resulting in a mean gray to white matter cerebral blood flow ratio of 2.67 +/- 0.07. This value is in excellent agreement with values published in the literature.