Aceptable day-to-day variability of cerebral perfusion using a non-invasive image derived input function in [15O]H2O PET

The PET radiotracer [ 15 O]H 2 O is the reference standard for quantifying myocardial blood flow and cerebral blood flow. However, the requirement for an invasive arterial input function limits clinical applicability. An alternative, non-invasive image-derived input function from the left ventricle has been validated for the heart, and recent work has shown that the image-derived input function obtained in an additional scan of the heart can be applied for the brain at rest and during acetazolamide-induced vasodilation. To further establish its clinical and research utility, we investigated the intra-individual day-to-day variability. Eight healthy participants underwent [ 15 O]H 2 O PET to assess cerebral blood flow on 2 occasions within 14 days using an image-derived input function placed in the left ventricle of the heart at rest. Day-to-day variability was assessed with Bland-Altman Plots, correlation coefficients, and Coefficient of Variation across three larger and three smaller brain regions. Day 1 and Day 2 measurements showed strong linear associations at rest (r = 0.75–0.85), during stress (r = 0.78–0.86) and for perfusion capacity (r = 0.75–0.82). Limits of agreement ranged from -0.096 to 0.121 mL/(g·min) at rest, -0.164 to 0.164 mL/(g·min) during stress and -0.255 to 0.189 mL/(g·min) for perfusion capacity. Coefficient of variation ranged from 11.9% to 12.6% at rest, 12.3% to 12.9% during stress and 7.8% to 8.2% for perfusion capacity. This non-invasive method is reproducible on a group level with coefficients of variation below 13% and is thus suitable for detecting group-level changes in intervention studies, although individual variability is substantial.