Quantitative, vascular T2 measurements are of interest for applications such as MR oximetry. In the situation of a vessel with long T2 relaxation times embedded in tissue with relatively short T2 values, contamination of the blood signal from the surrounding tissue can bias T2 measurements. Limited data sampling and vessels running obliquely through the imaging slice can cause significant signal contamination. Using a model of these effects to predict the behavior of T2 measurements under a range of conditions, a set of parameters that provides the best combination of measurement accuracy with a signal-to-noise ratio as high as possible is proposed.