Repeatability of Automated Global and Local Arterial Input Function Deconvolution Methods for Generating Cerebral Blood Flow Maps

Aleksandra Maria Stankiewicz^1,2, Ona Wu², Thomas Benner², Robert E. Irie², Tracy T. Batchelor², A Gregory Sorensen²

Perfusion-weighted Magnetic Resonance (MR) Imaging is used to assess the risk of tissue infarction in acute stroke patients and tumor angiogenesis in cancer patients. We compared circular global arterial input function (AIF) and local AIF algorithms, recently proposed automated methods for MR signal deconvolution. 13 patients with 2 MR scans within 48 hours were studied. The variation between global AIF cerebral blood flow (CBF) maps from the first and second scans was 0.220 0.043, and between local AIF CBF maps was 0.263 0.041 (P-value = 0.0015). Superior repeatability of global AIF-based CBF maps may be important in speedy diagnosis and risk stratification.