Accurate and fast quantification of myocardial blood flow (MBF) with MR first-pass perfusion imaging techniques on a pixel-bypixel basis remains difficult due to relatively long calculation times and noise-sensitive algorithms. In this study, Zierler's central volume principle was used to develop an algorithm for the calculation of MBF with few assumptions on the shapes of residue curves. Simulation was performed to evaluate the accuracy of this algorithm in the determination of MBF. To examine our algorithm in vivo, studies were performed in nine normal dogs. Two first-pass perfusion imaging sessions were performed with the administration of the intravascular contrast agent Gadomer at rest and during dipyridamole-induced vasodilation. Radiolabeled microspheres were injected to measure MBF at the same time. MBF measurements in dogs using MR methods correlated well with the microsphere measurements (R 2 ؍ 0.96, slope ؍ 0.9), demonstrating a fair accuracy in the perfusion measurements at rest and during the vasodilation stress. In addition to its accuracy, this method can also be optimized to run relatively fast, providing potential for fast and Quantification of myocardial blood flow (MBF) has been shown to be an effective tool for diagnosing blood flow defects (regional or global myocardium) and monitoring the effectiveness of therapeutic treatment (1-5). In particular, the application of first-pass techniques to each pixel of an image to produce an accurate blood flow map allows visualization of regional differences in blood flow with relatively high resolution, and is a noninvasive approach of assessing the severity of coronary artery blockage (6 -9).MBF is quantified by deconvolving tissue residue curves measured by dynamic first-pass images and by finding the peak of the resulting impulse response. Consequently, the accuracy of a first-pass algorithm depends largely on its ability to represent a wide variety of impulse response curves. For this reason, model-independent algorithms (10,11), which make few assumptions about the shape of the impulse response, are advantageous. However, because model-independent algorithms require curve-fitting of many parameters, they require special techniques to control noise susceptibility (10). One common way of stabilizing these methods is to introduce a set of smoothing constraints with a weight that depends on the noise level of the data. This "regularization method" is similar to applying a low-pass filter (10). While this technique substantially improves the conditioning of the deconvolution, it increases computation time and strong regularization can lead to underestimation of parameters to be measured.In this study we investigated a new model-independent technique that can be performed with relatively few parameters and does not require regularization. Like other quantification methods it utilizes a deconvolution based on Zierler's central volume principle. However, by choosing a simple representation of the impulse response curve we can achieve low noise sensiti...