Quantitative analysis of Gd-DTPA enhanced dynamic MR images has potential for discriminating lesions, especially because the introduction of clinical fast imaging techniques has enabled good sampling of the relatively rapid Gd-DTPA wash-in curves. Analysis of such data requires curve fitting to a nonlinear model, which to date has been performed using a nonlinear least squares (NLLS) fitting procedure. However, this method often fails to converge to the appropriate minima without good initial parameter estimates when multi-exponential models are involved, making automated analysis of complete multislice or volume data sets problematic. In this report we demonstrate the robust performance of a simplex minimization procedure compared with NLLS, by the method of Marquardt, using a Monte Carlo simulation. Further, we illustrate the applicability of such a technique to the analysis of dynamic contrast enhanced images on a pixel-by-pixel basis. As a preliminary example, the technique is applied to a breast lesion but is expected to be suitable for examination of many lesion types.