Spatio-temporal deconvolution of perfusion CT data in rectal tumor patients

Anatomical imaging modalities such as MR and CT produce high resolution images that are essential in modern health care. Anatomical imaging, however, can be supplemented with functional imaging to provide additional insights into the status and progression of disease. In perfusion imaging, a bolus of contrast entering a tissue region of interest is carefully monitored in order to derive hemodynamic parameters. While useful, perfusion parameter maps tend to be noisy, in part because the parameters of every voxel are computed independently of its neighboring voxels. A spatio-temporally correlated approach to estimating perfusion parameters has been previously introduced. In this paper, we expand upon this earlier formulation and demonstrate its functionality on a set of rectal tumor patients undergoing CT perfusion (CTp) studies. Our results show that our spatio-temporal deconvolution approach is less noisy than conventional singular value deconvolution approach, and show greater separation between normal and cancerous regions.