ABSTRACT
In dynamic susceptibility contrast-enhanced magnetic resonance imaging (DSC-MRI), an arterial input function (AIF) is usually obtained from a time-concentration curve (TCC) of the cerebral artery. This study was aimed at developing an alternative technique for reconstructing AIF from TCCs of multiple brain regions. AIF was formulated by a multi-exponential function using four parameters, and the parameters were determined so that the AIF curves convolved with a model of tissue response reproduced the measured TCCs for 20 regions. Systematic simulations were performed to evaluate the effects of possible error sources. DSC-MRI and positron emission tomography (PET) studies were performed on 14 patients with major cerebral artery occlusion. Cerebral blood flow (CBF) images were calculated from DSC-MRI data, using our novel method alongside conventional AIF estimations, and compared with those from (15)O-PET. Simulations showed that the calculated CBF values were sensitive to variations in the assumptions regarding cerebral blood volume. Nevertheless, AIFs were reasonably reconstructed for all patients. The difference in CBF values between DSC-MRI and PET was -2.2 ± 7.4 ml/100 g/min (r = 0.55, p < 0.01) for our method, versus -0.2 ± 8.2 ml/100 g/min (r = 0.47, p = 0.01) for the conventional method. The difference in the ratio of affected to unaffected hemispheres between DSC-MRI and PET was 0.07 ± 0.09 (r = 0.82, p < 0.01) for our method, versus 0.07 ± 0.09 (r = 0.83, p < 0.01) for the conventional method. The contrasts in CBF images from our method were the same as those from the conventional method. These findings suggest the feasibility of assessing CBF without arterial blood signals.
