Automatic Image-Derived Estimation of the Arterial Whole-Blood Input Function from Dynamic Cerebral PET with $$^{18}$$F-Choline

The arterial concentration of the radiopharmaceutical \({}^{18}\text {F}\)-choline is needed to estimate its absorption by tumors and other tissues. The blood concentration of \({}^{18}\text {F}\)-choline changes as it interacts with tissues, and so it is represented as a function with respect to time, the so-called Input Function (IF). In this paper, we present the estimation of an arterial whole-blood Image-Derived Input Function (IDIF) from the PET image, needed to model its absorption. The sagittal and transverse brain venous sinuses are automatically segmented based on the top-hat morphological transform. Such segmentation provides an estimation of the venous whole-blood IDIF. It is then corrected to obtain the arterial whole-blood IDIF by relating the amount of radioactivity material entering the brain region with the amount leaving it and the amount remaining. We compare the automatic venous whole-blood IDIF with a whole-blood venous IDIF from a region manually segmented. Also, we compare the automatic arterial whole-blood IDIF with the arterial IF obtained with serial blood samples on the radial artery. Quantitative measures indicate the overall accuracy of the estimation.