Reference Cluster Normalization Improves Detection of Frontotemporal Lobar Degeneration by Means of FDG-PET

Citation metadata

From: PLoS ONE(Vol. 8, Issue 2)
Publisher: Public Library of Science
Document Type: Article
Length: 5,830 words
Lexile Measure: 1520L

Document controls

Main content

Article Preview :

Author(s): Juergen Dukart 1 , 3 , * , Robert Perneczky 6 , 8 , 12 , Stefan Förster 9 , Henryk Barthel 4 , 7 , Janine Diehl-Schmid 6 , 8 , Bogdan Draganski 1 , 5 , Hellmuth Obrig 2 , 3 , Emiliano Santarnecchi 10 , Alexander Drzezga 9 , Andreas Fellgiebel 11 , Richard Frackowiak 1 , Alexander Kurz 6 , 8 , Karsten Müller 3 , Osama Sabri 4 , 7 , Matthias L. Schroeter 2 , 3 , 4 , 6 , Igor Yakushev 9 , 11

Introduction

In brain imaging by means of positron emission tomography (PET) and single-photon emission computed tomography (SPECT), scaling of tracer uptake to a reference region is in most cases essential for analyses of non-quantitative data. An ideal reference region should not be affected by brain pathology and should be easy to image/analyse. The choice of the appropriate reference region is especially problematic in subjects with neurodegenerative disorders who show early metabolic and perfusion deficits [1]-[5].

Using statistical parametric mapping (SPM), we have recently proposed a data-driven method for normalization of [18F] fluorodeoxyglucose (FDG) uptake in cases with preclinical and manifest Alzheimer's disease (AD) dementia [6]. As compared with traditional intensity normalization to an a priori defined reference region, the reference cluster (RC) is defined by a contrast showing areas with increased activity in patients relative to controls after global mean normalization. This RC approach proved to detect AD-related hypometabolism in a more sensitive manner than traditional ROI-based normalizations [6]. Originally developed on clinical image data, the method was soon validated in simulation studies and extended to perfusion studies in Parkinson's disease (PD), another common neurodegenerative disorder [7]-[9].

To be applicable in clinical settings, any diagnostic test must be robust to variability in clinical presentation/assessments and methodological factors. To this end, we performed a bi-central study with a cross-validation design in frontotemporal lobar degeneration (FTLD). As compared to AD and PD, FTLD is characterized by a substantially higher heterogeneity in respect to histopathological, clinical, and imaging presentation [10]-[12]. FDG-PET plays a well-established role in assisting early detection and differentiation of this severe neurodegenerative disorder [13]-[25], after AD the second most common cause of presenile dementia [26]. Furthermore, as compared to previous single-center applications of the RC normalization in AD and PD [6], [27], [28], here we assess performance of the method in cross-center settings. I.e., a RC obtained at one center is applied for data normalization from another center. Results of this data-driven approach are then compared with normalization to common reference regions.

In this work, we examine the impact of intensity normalization to cerebellum (CBL), primary sensorimotor cortex (SMC), cerebral global mean (CGM) and to RC on the accuracy of FDG-PET to detect FTLD-specific metabolic deficits and to discriminate between patients with mild FTLD and healthy subjects.

Methods

Subjects

Patients were retrospectively identified from a database of subjects from the memory clinic of the Department of Psychiatry at the Technische Universität München (hereafter referred to as center...

Source Citation

Source Citation   

Gale Document Number: GALE|A478882269