Feasibility of pharmacokinetic parametric PET images in scaled subprofile modelling using principal component analysis

Débora E Peretti,Remco J Renken,Fransje E Reesink,Bauke M De Jong,Peter P De Deyn,Rudi A.J.O Dierckx,Janine Doorduin,Ronald Boellaard,David Vállez García

doi:10.1016/j.nicl.2021.102625

Débora E Peretti, Remco J Renken + Show 7 more

Open Access

https://doi.org/10.1016/j.nicl.2021.102625

Copy DOI

Abstract

Scaled subprofile model using principal component analysis (SSM/PCA) is a multivariate analysis technique used, mainly in [18F]-2-fluoro-2-deoxy-d-glucose (FDG) PET studies, for the generation of disease-specific metabolic patterns (DP) that may aid with the classification of subjects with neurological disorders, like Alzheimer’s disease (AD). The aim of this study was to explore the feasibility of using quantitative parametric images for this type of analysis, with dynamic [11C]-labelled Pittsburgh Compound B (PIB) PET data as an example. Therefore, 15 AD patients and 15 healthy control subjects were included in an SSM/PCA analysis to generate four AD-DPs using relative cerebral blood flow (R1), binding potential (BPND) and SUVR images derived from dynamic PIB and static FDG-PET studies. Furthermore, 49 new subjects with a variety of neurodegenerative cognitive disorders were tested against these DPs. The AD-DP was characterized by a reduction in the frontal, parietal, and temporal lobes voxel values for R1 and SUVR-FDG DPs; and by a general increase of values in cortical areas for BPND and SUVR-PIB DPs. In conclusion, the results suggest that the combination of parametric images derived from a single dynamic scan might be a good alternative for subject classification instead of using 2 independent PET studies.

Highlights

Proper interpretation of positron emission tomography (PET) scans is important for clinical diagnosis, and to monitor disease progression and response to treatment (Lammertsma, 2017)
This study provides the first step for future research studies using parametric images derived from pharmacokinetic modelling in Scaled subprofile model using principal component analysis (SSM/PCA) in clinical settings
disease-specific metabolic patterns (DP) voxels with negative values indicated regions where there was a decrease in the parameter (e.g. R1 or SUVR) for Alzheimer’s disease (AD) patients compared to the Healthy control (HC) group and vice versa for the positive values

Summary

Introduction

Proper interpretation of positron emission tomography (PET) scans is important for clinical diagnosis, and to monitor disease progression and response to treatment (Lammertsma, 2017) This interpretation is often made through visual inspection of the images or by means of semiquantitative approaches such as standardized uptake values (SUV) or a ratio (SUVR), when there is a reference region without specific binding of the tracer. To obtain an optimal quantification of the (patho)physiology under study, it is necessary to decompose the PET signal into its different components, or kinetic ‘states’ (Carson, 2003), for example, in a compartment that expresses tracer concentration that is bound to the target and a separate compartment with free tracer in tissue (Gunn et al, 2001) These quantitative metrics can be obtained by applying pharmacokinetic modelling to PET data. Phar macokinetic models can be applied to the whole PET dataset at a voxellevel, resulting in high-quality parametric images that can be used to perform a visual assessment, with the potential to reduce misclassifi cation and improve the inter-reader agreement, and to accurately quantify tracer uptake (Collij et al, 2019; Lammertsma, 2017; Peretti et al, 2019c)

Objectives

Methods

Results

Conclusion