Elsevier

NeuroImage

Volume 48, Issue 2, 1 November 2009, Pages 329-338
NeuroImage

Strategies for the generation of parametric images of [11C]PIB with plasma input functions considering discriminations and reproducibility

https://doi.org/10.1016/j.neuroimage.2009.06.079Get rights and content

Abstract

Pittsburgh compound B or [11C]PIB is an amyloid imaging agent which shows a clear differentiation between subjects with Alzheimer's disease (AD) and controls. However the observed signal difference in other forms of dementia such as dementia with Lewy bodies (DLB) is smaller, and mild cognitively impaired (MCI) subjects and some healthy elderly normals may show intermediate levels of [11C]PIB binding. The cerebellum, a commonly used reference region for non-specific tracer uptake in [11C]PIB studies in AD may not be valid in Prion disorders or monogenic forms of AD. The aim of this work was to: 1—compare methods for generating parametric maps of [11C]PIB retention in tissue using a plasma input function in respect of their ability to discriminate between AD subjects and controls and 2—estimate the test–retest reproducibility in AD subjects. 12  AD subjects (5 of which underwent a repeat scan within 6 weeks) and 10 control subjects had 90 minute [11C]PIB dynamic PET scans, and arterial plasma input functions were measured. Parametric maps were generated with graphical analysis of reversible binding (Logan plot), irreversible binding (Patlak plot), and spectral analysis. Between group differentiation was calculated using Student's t-test and comparisons between different methods were made using p values. Reproducibility was assessed by intraclass correlation coefficients (ICC). We found that the 75 min value of the impulse response function showed the best group differentiation and had a higher ICC than volume of distribution maps generated from Logan and spectral analysis. Patlak analysis of [11C]PIB binding was the least reproducible.

Introduction

Alzheimer's disease (AD) is characterised by the demonstration of amyloid plaques and neurofibrillary tangles in the brain. The clinical diagnosis of AD is based on the NINCDS-ADRDA (McKhann et al., 1984) and DSM 1V criteria. With the introduction of [11C]PIB or Pittsburgh compound B, it is now possible to detect amyloid plaque deposition in vivo with Positron Emission Tomography (PET) in subjects with AD. Modification of the amyloid binding histological dye, thioflavin-T led to the finding that neutral benzothiazoles also bind to amyloid with high affinity and additionally cross the blood–brain barrier (Klunk et al., 2001). Specifically, the benzothiazole amyloid binding agent 2-(4′ methylaminophenyl) benzothiazole (called BTA-1) and related compounds bind to amyloid-beta (Aβ) peptide fibrils with low nanomolar affinity, enter the brain in amounts sufficient for imaging with PET and clear rapidly from normal brain tissue (Mathis et al., 2002, Mathis et al., 2003, Wang et al., 2002). At the low nanomolar concentrations typically used in PET studies, BTA-1 binds to amyloid plaques in post mortem brain slices but not to intracellular neurofibrillary tangles. In vitro studies suggest that, while BTA-1 binds to fibrillar Aβ deposits found in the cortex and striatum, it does not bind to amorphous Aβ deposits found in cerebellum (Mathis et al., 2003).

The first PET imaging study in humans showed that [11C]PIB standard uptake values (SUV) were increased twofold in cortical association areas of AD patients compared to healthy control subjects (Klunk et al., 2004). The first tracer kinetic modelling report on [11C]PIB binding in humans concluded that graphical analysis of reversible binding (Logan et al., 1990) using a plasma input function was the method-of-choice providing stable region-of-interest results and robust parametric images (Price et al., 2005). Subsequently, the same laboratory reported that a 90 min target-to-cerebellum [11C]PIB uptake ratio produced the largest effect size while graphical analysis of reversible binding (Logan et al., 1996) using a cerebellar time-activity curve as the input function produced the most reproducible distribution volume ratios (Lopresti et al., 2005).

Several authors have now studied the generation of parametric maps of [11C]PIB binding with simplified methods, i.e. not requiring a plasma input function, for the discrimination between healthy controls and subjects with AD or mild cognitive impairment (MCI) (Mikhno et al., 2008, Ng et al., 2007, Ziolko et al., 2006, Zhou et al., 2007). In a recent paper, Yaqub et al. (2008) presented an exhaustive comparison of simplified parametric methods for the analysis of [11C]PIB studies and concluded that the modified simplified reference tissue model (Wu and Carson, 2002) was the method of choice.

The evaluation of [11C]PIB kinetics in many of the approaches employed in the above listed publications is based on the apparent reversibility of [11C]PIB binding to Aβ over the time course of a 90 min PET scan. However, Bacskai et al. (2003) found in a multiphoton imaging study in transgenic mice that PIB remained associated with amyloid deposits for several days. Based on analysis of data from the first [11C]PIB scans in humans which were acquired over 60 min, Blomquist proposed the use of an irreversible graphical model for the quantification of tracer uptake (Patlak et al., 1983) to generate parametric maps of [11C]PIB accumulation (Blomquist et al., 2005, Blomquist et al., 2008, Hinz et al., 2005b).

Another important aspect in the analysis of [11C]PIB binding is the selection of a reference region representing non-specific signal such as the cerebellum, white matter or pons. While a cerebellar reference may be valid for sporadic Alzheimer's disease, pathologies such as Prion disease can result in a significant amount of amyloid in the cerebellum (Mead et al., 2006, Watanabe and Duchen, 1993). A recent [11C]PIB PET report has also suggested that there is significant amyloid deposition in familial AD (Kaufer et al., 2008). In these conditions, finding a suitable reference region could potentially be difficult. A reference region should always be validated as being effectively void of specific binding in proportion to the specific binding in the target area before using it for the quantification of imaging studies (Asselin et al., 2007). As more mild cognitive impairment patients and patients with subjective memory impairment are studied, and various longitudinal studies are examined, it is important to use analysis methods with maximum discriminative power and to establish the reproducibility of the methods of analysis.

In this study we used three of the most common kinetic modelling methods employing an arterial plasma input function to generate parametric maps of [11C]PIB binding (Logan plot, Patlak plot, spectral analysis) and compared their test–retest reproducibility in a group of AD subjects and their ability to discriminate between a group of AD subjects and controls.

Section snippets

Subjects

For the discriminative part of the study, 12 sporadic AD subjects and 10 age and gender matched control subjects were selected (Table 1). In order to examine test–retest reproducibility 5 of these AD subjects underwent a repeat [11C]PIB scan within an interval of less than 6 weeks. These [11C]PIB scans were acquired as a part of a previously published larger research study comparing [18F]FDG and [11C]PIB PET in AD (Edison et al., 2007). All subjects were recruited from the Imperial College

Input function

In Fig. 1, the observed time courses of the ratio of plasma activity concentration over whole blood activity concentration (POB ratio) from the 10 discrete arterial blood samples are plotted for the two cohorts. Immediately after radiotracer injection the POB ratio quickly rises to reach a plateau at a nearly constant value of about 1.45. The two cohorts did not statistically differ in their POB time courses. Such time courses of POB ratios are characteristic for radiotracers that passively

Discussion

Our measurements of the parent fraction (Fig. 3) appear to be in reasonable agreement with the previously published results by Price et al. (2005). This is of some importance as the fraction of radioactivity in arterial plasma due to unmetabolised [11C]PIB decreases rapidly. The mean parent fraction of [11C]PIB was found to be about 75% after 5 min, 28% after 15 min, 10% after 40 min and less than 5% after 90 min scan time in our study of 22 subjects. Due to difficulties with the sensitivity of

Acknowledgments

We gratefully acknowledge the excellent work of Safiye Osman with the bioanalysis and quality control teams. We would like to particularly thank Andy Blythe and Dave Turton for their efforts in characterising and minimising the stickiness properties of [11C]PIB. Leonhard Schnorr, Hope McDevitt, and Andreanna Williams receive our thanks for their making the scans possible. We are very grateful to Prof. Martin Rossor for his help with the recruitment of Alzheimer's disease subjects for this

References (43)

  • BlomquistG. et al.

    Unidirectional influx and net accumulation of PIB

    The Open Neuroimaging Journal

    (2008)
  • BrixG. et al.

    Performance evaluation of a whole-body PET scanner using the NEMA protocol. National Electrical Manufacturers Association

    J. Nucl. Med.

    (1997)
  • CollignonA. et al.

    Automated multi-modality image registration based on information theory

  • CunninghamV.J. et al.

    Spectral analysis of dynamic PET studies

    J. Cereb. Blood Flow Metab.

    (1993)
  • DefriseM. et al.

    Exact and approximate rebinning algorithms for 3-D PET data

    IEEE Trans. Med. Imaging

    (1997)
  • EdisonP. et al.

    Amyloid, hypometabolism, and cognition in Alzheimer disease: an [11C]PIB and [18F]FDG PET study

    Neurology

    (2007)
  • HammersA. et al.

    Statistical neuroanatomy of the human inferior frontal gyrus and probabilistic atlas in a standard stereotaxic space

    Hum. Brain Mapp.

    (2007)
  • HinzR. et al.

    Determination of tracer arrival delay with spectral analysis

    IEEE Trans. Nucl. Sci.

    (2006)
  • HinzR. et al.

    Parametric imaging of [11C]PIB studies using spectral analysis

    J Cereb Blood Flow Metab

    (2005)
  • HinzR. et al.

    Simplified analysis of [11C]PIB images in Alzheimer's disease patients and control subjects

    J Nucl Med

    (2005)
  • KauferI.D. et al.

    Cerebellar amyloid in a case of early-onset dementia with a presenilin-1 mutation: correlation to clinical phenotype

    Alzheimer's and Dementia

    (2008)
  • Cited by (0)

    View full text