Variability of the arterial input function in small-animal dynamic PET imaging

Abstract

Dynamic positron emission tomography (PET) combined with tracer kinetic modeling enables non-invasive quantification of biochemical processes. A prerequisite is availability of the arterial input function (AIF), which, in small-animal PET imaging, involves labor-intensive terminal surgery. Deep learning based input function (DLIF) allows estimation of whole-blood tracer concentration from PET data, avoiding arterial cannulation, but require comprehensive validation. The aim of this study was to collect dynamic PET and AIF data under controlled conditions and to evaluate the variability of AIF, image-derived input function (IDIF), and kinetic modeling parameters, which is important for future DLIF model training. Dynamic PET and AIF data were collected prospectively from 112 mice in groups with varying experimental conditions, including radiotracer injection volume, injection time, withdrawal rate, mouse age, strain, radiopharmaceutical, and PET scanner. Brain, myocardium, left ventricle and liver were delineated for kinetic modeling and IDIF generation. Curve features and kinetic modeling parameters were computed, using both AIF and IDIF, and compared across groups using box plots and statistical tests. Intra-subject repeatability was evaluated in six mice using three small-volume radiotracer injections. Experimental factors such as mouse strain, injection time, withdrawal rate, PET scanner and radiopharmaceutical significantly affect AIF and IDIF shapes, while injection volume and mouse age, did not introduce bias. AIF measurements within the same subject were highly repeatable. This study collected a comprehensive dataset of dynamic PET and AIF measurements under controlled conditions to evaluate the variability of AIF, IDIF, and kinetic modeling parameters. The findings provide valuable insights into input function variability, with potential implications for the future development of DLIF models across diverse experimental conditions.

Ähnliche Arbeiten