Dynamic functional imaging techniques promise powerful tools for the visualization and elucidation of important disease-causing physiologic processes in living tissue. Most applications aim to find temporal-spatial patterns associated with different disease stages. When multiple functional biomarkers are targeted, imagery signals often represent a composite of more than one distinct functional source independent of spatial resolution. Here we introduce a hybrid blind source separation method that is able to factorize underlying source images and time activity curves from dynamically mixed image sequence. The algorithm is based on a compartment latent variable model, whose parameters are estimated using multivariate clustering and/or independent component analysis. We demonstrate the principle of the approach on tumor microvascular characterization using dynamic contrast-enhanced magnetic resonance imaging for monitoring response to anti-angiogenic therapies.