Automatic measurements of arterial input and venous output functions on cerebral computed tomography perfusion images: A preliminary study
A new single-step method is described for estimation of local cerebral blood volume (CBV) from multiple rapidly acquired T2- or T2*-weighted MRI images after bolus administration of a susceptibility contrast agent. This method improves on existing methods in three ways. First, the method includes the baseline scan intensity value as a fitted parameter. Second, the model is fitted directly to the original scan intensity values instead of to transformed concentration curves, avoiding the propagation of errors that occurs in the transformation. Third, the equations are reparameterized with CBV fitted directly as a model parameter. Parameter estimation methods are compared by Monte Carlo simulation. The direct method described here yields more precise parameter estimates, with smaller mean absolute deviations from the true parameter values, than the existing method when compared using simulated data. Implementation is discussed and numerical evaluation of the Digamma or Psi function is described.