Development and Validation of an Accurate Input Function from Carotid Arteries using the uEXPLORER

Abstract

For a dedicated brain scan, the carotid artery is the best location for acquiring an image-based input function. With improvements in PET spatial resolution, accurate quantitation may be achieved with PET data alone. With the ability to cover both the carotid artery and the thorax at high spatial resolution, the uEXPLORER datasets provide a unique opportunity to develop and validate input functions in multiple regions such as the carotid artery. The regions containing the carotid arteries were first manually identified using reconstructed images consisting of the first 60 seconds of data post-injection. The image-based point spread function (PSF) was measured using off-center phantom scans to approximate the locations of the carotid artery. The same reconstruction approach was used for both the phantom scans and the volunteer scans. The structure of the carotid artery at each slice was generated using a deconvolution approach. An additional constraint of a uniform activity distribution within the carotid artery was added in the deconvolution approach. The acquired carotid artery structure was then applied to the dynamic frames (1-hour data) from volunteer scans for partial volume correction to acquire the input function (CA-IF). The input function from the descending aorta (DA-IF) was also extracted as a gold standard. The area-under-curve (AUC) ratio between the two input functions was used to evaluate the accuracy of the method. Without correction, there was a significant visual difference between CA-IF and the DA-IF, which was reduced dramatically after correction. The quantitation difference was dramatically reduced with the proposed correction method. The AUC ratio between the two input functions was 0.78+-0.04 (original), and was 1.00+-0.03 after correction, suggesting much improved quantitative accuracy. The results demonstrated that with improved image resolution and sensitivity, it is possible to accurately acquire the input function from carotid arteries without reliance on extra anatomical imaging approaches such as MRI.