Hemodynamic segmentation of MR perfusion images in patients with unilateral carotid stenosis using independent component analysis.

PURPOSE To assess spatiotemporal patterns of local perfusion in MR brain perfusion images for patients with unilateral carotid stenosis. MATERIALS AND METHODS MR perfusion images acquired from 12 patients with unilateral carotid stenosis were retrospectively segmented using independent component analysis. Regions of interest selected from the segmentation results were used to measure the time to peak (TTP), time to maximum of the residue curve (Tmax), and mean transit time (MTT) for arteries and brain parenchyma on normal and stenotic sides. RESULTS The tissue types that can be identified in segmentation results include normal artery, artery with proximal stenosis, normal brain parenchyma, delayed-perfusion brain parenchyma, vein and sinus, choroid plexus, cerebrospinal fluid, and leukoaraiosis. Concentration-time curves measured on normal and stenotic artery regions can be modeled as two arterial input functions. Unilateral carotid stenosis associated with either middle cerebral artery stenosis or poor collateral circulation at the circle of Willis resulted in increased differences of brain parenchyma TTP, Tmax, and MTT between normal and stenotic sides. CONCLUSION Independent component analysis on perfusion images of patients with unilateral carotid stenosis provides useful spatiotemporal information regarding blood supply to arteries and local perfusion of brain parenchyma.