Vessel size imaging is a new method that is based on simultaneous measurement of the changes R 2 and R* 2 in relaxation rate constants induced by the injection of an intravascular su-perparamagnetic contrast agent. Using the static dephasing approximation for R* 2 estimation and the slow-diffusion approximation for R 2 estimation, it is shown that the ratio R 2 / R* 2 can be expressed as a function of the susceptibility difference between vessels and brain tissue, the brain water diffusion coefficient, and a weighted mean of vessel sizes. Comparison of the results with 1) the Monte Carlo simulations used to quantify the relationship between tissue parameters and susceptibility contrast, 2) the experimental MRI data in the normal rat brain, and 3) the histologic data establishes the validity of this approach. This technique, which allows images of a weighted mean of the vessel size to be obtained, could be useful for in vivo studies of tumor vascularization. Magn Re-son Med 45:397-408, 2001. Magnetic resonance imaging (MRI) gives access to information on cerebral perfusion, which is of importance for the diagnosis and therapeutic follow-up of various pathol-ogies. Accurate mapping of cerebral blood volume (CBV), cerebral blood flow (CBF), and mean transit time (MTT) is an area of intense research, and CBV-, CBF-and MTT-weighted imaging are beginning to be used in clinical practice. Large vessels can be imaged; however, information on microvascular architecture is still hardly accessible. Information on the vascular component of tissues may be obtained by various contrast-enhanced MR methods. Compared to bolus tracking techniques (1), steady-state methods offer the potential of a higher signal-to-noise ratio (SNR), and thus a higher spatial resolution. One kind of steady-state approach for CBV imaging is based on the measurement, in T 1-weighted MR images, of changes in signal intensity after injection of a contrast agent (2). This method assumes that brain tissue can be modeled as two nonexchanging compartments: intra-and extravascular. An entirely different steady-state T 1 technique for CBV mapping is based on the analysis of changes in tissue T 1 due to exchange of water between the two compartments (3). Finally, another class of steady-state CBV imaging techniques relies on susceptibility-induced contrast (4). Susceptibility contrast imaging shows changes in signal intensity related to magnetic susceptibility differences between the intra-and extravascular compartments. This phenomenon underlies the blood oxygen level-dependent (BOLD) contrast (5), which is the basis of functional MRI. A similar effect is obtained with exogenous paramagnetic or superparamagnetic contrast agents, which increase the magnetic susceptibility differences between blood vessels and surrounding tissues. The induced long-range magnetic field perturbations extend to adjacent tissues and increase the transverse relaxation rate constants R 2 and R* 2. Two phenomena affect the transverse relaxation rates of tissue water in the presence of a contrast agent in the blood pool (6). First, magnetic field perturbations increase the heterogeneity of the phase distribution across the voxel. Reversible spin dephasing occurs with associated signal loss in gradient-echo experiments (GE), resulting in an increase R* 2 in the relaxation rate. However, in the absence of diffusional motion, R 2 1/T 2 remains unchanged. Second, diffusion of water molecules in magnetic field gradients introduces two competitive effects: 1) irreversible losses of phase coherence and signal attenua-tion in spin echo (SE) experiments; and 2) for rapid spatial variations of magnetic field, possible averaging of phase differences by motional narrowing, resulting in reduced T* 2 and T 2 changes. A salient implication of these effects is that changes R 2 and R* 2 in the transverse relaxation rates depend on the size and architecture of the vascular compartment. One topic of great interest in susceptibility contrast imaging concerns the sensitivity of MRI experiments to the distribution of vessel sizes. On the basis of Monte Carlo (MC) simulations, Boxerman et al. (4) suggested that information about microvascularization could be obtained by measuring, at various TEs, the change R 2 induced by injection of an intravascular contrast agent. By combining SE and GE BOLD MRI, Prinster et al. (7) demonstrated that the ratio R* 2 /R 2 is related to the properties of the vascu-lar environment. Dennie and coworkers (8) used a similar technique to show that the ratio R* 2 /R 2 could be used as an index of the mean vessel size to compare normal and tumor tissues in the rat brain. To obtain information on the vessel size distribution they compared the experimental R* 2 /R 2 ratio to that obtained from MC simulations, which is a time-consuming technique. A vessel size index deduced analytically from the measured changes R 2 and R* 2 in relaxation rates would be desirable. A relation between R* 2 and the susceptibility difference between blood and tissue has been proposed (6), and has been successfully used to obtain information on the saturation in oxygen of cerebral venous blood (9). More recently, Kiselev and Posse (10,11) used a deterministic analytical method to establish a relationship between R 2 and , a