Quantitative Image-Derived Input Function for the Estimation of Cerebral Blood Flow

Project by Thomas Lund Anderesen

Introduction

The quantification of cerebral blood flow using [15O]H2O PET relies on precise input function estimation, conventionally obtained via arterial blood sampling. This study investigates an automated approach to image-derived input function (IDIF) extraction using a long axial field-of-view (LAFOV) PET/CT scanner, with the aim of replacing invasive procedures.

Project Background

The LAFOV PET/CT scanner enables a direct and non-invasive sampling of the blood activity concentration by extracting the arterial input function (AIF) directly from the PET image data. In this project the extracted IDIF was compared with the AIF obtained through simultaneous arterial blood sampling. Modelling of regional cerebral blood flow (rCBF) using kinetic compartment modelling with [15O]H2O PET data using IDIF and AIF was compared using the two blood input functions.

Project Implementation

The IDIF closely matched the gold standard across multiple metrics, including peak shape and area under the curve.

Cerebral blood flow estimates based on IDIF offered a reliable alternative to invasive blood sampling, as the input function can be measured directly from the PET image data. The voxel-wise perfusion maps (Figure 3) with quantitative differences between AIF- and IDIF-based estimates were minimal across the brain. Ultimately, the extraction of the blood concentration function has been implemented clinically for the benefit for patients.

P13_fig3
Figure 3. Perfusion maps based on AIF and IDIF show near-identical results. The voxel-wise difference (right) confirms minimal variation between the two methods.
P13_fig1
Figure 1. Brain TAC modelled using both IDIF (green) and AIF (red), showing close alignment. Estimated parameters (K₁, Vd) are in strong agreement, while differences in delay (ΔT) and dispersion (τ) highlight technical artefacts avoided with IDIF.
Figure 2: Results of Low-Activity Scanning
Figure 2. Raw IDIF (dashed green) shows an earlier peak than AIF (red) due to the absence of dispersion and delay. After correction (solid green), the curves align closely, confirming that differences are technical in origin.

Contact Information

Name: Thomas Lund Andersen
Email: thomas.lund.andersen@regionh.dk
Location: Department of Clinical Physiology and Nuclear Medicine, Rigshospitalet, Denmark
Position: Physicist

Publications

Automated Quantitative Image-Derived Input Function for the Estimation of Cerebral Blood Flow Using Oxygen-15-Labelled Water on a Long-Axial Field-of-View PET/CT Scanner