Imaging of the vasculature and its functioning over the entire lesion may significantly aid in cancer diagnosis, assessment of prognosis, and therapeutic evaluation. In the current study we present a dynamic three-dimensional deuterium magnetic resonance imaging method that determines the intravascular volume fraction and water perfusion rate at a resolution of 2 mm(2)/pixel. The method was tested and utilized to characterize the vasculature of orthotopic MCF7 human breast cancer tumors in CD1-NU athymic mice. A new algorithm based on Patlak's kinetic model was developed to analyze the dynamic images acquired during and after termination of infusion with deuterated water. The resulting parametric maps spanned a wide range from 0.4 to 35.2% for the intravascular volume fraction and from 4 x 10(-6) to 3.9 x 10(-3) min(-1) for the perfusion rate and exhibited high intratumoral and intertumoral heterogeneity at both parameters. The intravascular volume fraction did not correlate with the corresponding perfusion rate, demonstrating the irregular outgrowth of tumor neovascularization. Averaging the data or analyzing at spatially degraded resolution completely masked the presence of both "hot spots" and hypoxic loci, highlighting the critical importance of high spatial resolution. The method is applicable to other types of tumors and animal models and may be extended to humans.