Dynamic contrast enhanced-MRI is essential for assessment of tissue microvasculature, and thus improved methods of parameter estimation for evaluation of tissue perfusion can be valuable. This work demonstrates how application of Variable Projection (VP) in perfusion analysis increases the robustness and computation efficiency compared to the conventional Levenberg-Marquardt (LM) method. A numeric simulation is used to validate the perfusion estimation and evaluate the robustness with respect to noise. The computation speed of VP is shown to be improved about 5× faster compared to LM method. VP-based perfusion analysis is demonstrated in eight patients with known prostate cancer.