Absolute quantification of cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time (NITT) are of great relevance for clinical applications. One of the widely used methods for quantification of these parameters is gamma-variate fitting. Traditional nonlinear regression methods for gamma-variate fitting are inaccurate and computationally demanding. In this study, we developed an adaptive total least square method (ATSSL) to fit a gamma-variate function to the delayed concentration-time course. For each concentration-time curve, the beginning and ending time point of the curve are adaptively determined online. After the curves were fitted, a robust method for automatically determination of arterial input function (AIF) from whole and region of interest (ROI) was developed. Using the obtained AIF and fitted gamma-variate concentration-time curve, the MTT, CBV, and CBF were calculated by utilizing singular value decomposition algorithm. Computer simulations show that the suggested method is adaptive, reliable, and insensitive to noise. Comparison with the traditional nonlinear regression method indicated that the presented method is more accurate and faster to determine the CBV, CBF and NITT. (C) 2003 Elsevier Inc. All rights reserved.