Chapter 36 - Absolute Cerebral Blood Flow with [ 15 O]Water and PET: Determination Without a Measured Input Function

This chapter presents a method to measure absolute cerebral blood flow cerebral blood flow (CBF) following a bolus injection using only dynamic tissue radioactivity values derived from reconstructed image data. As in all CBF methods, the input function is assumed to be common to all brain pixels. Each pixel's time–activity curve therefore has information about the local blood flow and the input function. By analyzing all pixels simultaneously, the input function can be estimated. Further, the theory and implementation of the new method and present results from simulated data and human studies are described. The Gauss–Newton (GN) method can produce accurate CBF measurements only when the measurements follow the assumed model. Tissue heterogeneity and other factors produce large bias in the GN flow values. These model errors produce much smaller biases in other algorithms that use a measured input function. By estimating both flow and the input function, the GN algorithm has more flexibility in minimizing Ψ and, therefore, is more sensitive to model errors. Two modified methods presented reduced the global flow bias, but eliminated the positive correlation of GN results with those using the measured input functions. Therefore, a better model for the distribution volume is required for the GN approach to produce accurate and reliable human CBF data.