A reference-based approach using whole-brain segmentation for image-derived PET input functions in CBF,OEF,and CMRO 2 quantification

Abstract

Quantitative measurements of cerebral blood flow (CBF), oxygen extraction fraction (OEF), and cerebral metabolic rate of oxygen (CMRO₂) using ¹⁵O-labeled PET generally require an arterial input function (AIF) obtained via invasive arterial blood sampling. This study aimed to develop a fully noninvasive method to reconstruct an image-derived input function (IDIF) directly from dynamic ¹⁵O₂–H₂¹⁵O PET data. Using a reference-based approach, data from 186 subjects with suspected cerebrovascular disease were analyzed. Segment-based reference values of CBF and OEF derived from 60 subjects were used to reconstruct IDIFs in the remaining 126 subjects through coefficient-of-variation weighted averaging of segmental tissue time–activity curves. The estimated IDIFs showed strong agreement with measured AIFs, yielding minimal bias (⩽0.005) and high correlations (r = 0.88, p < 0.001) for CBF, OEF, and CMRO₂. No significant differences were observed between IDIF- and AIF-based quantitative values. However, the reconstructed IDIFs exhibited mild regression-to-the-mean behavior, likely due to temporal smoothing. These findings demonstrate that input functions for labeled water and oxygen can be reliably estimated from tissue curves without blood sampling. The proposed method enables fully noninvasive quantification of CBF, OEF, and CMRO₂, supporting its clinical feasibility for ¹⁵O PET studies, but not for quantitative parameter estimation under stimulation.