Elsevier

NeuroImage

Volume 37, Issue 1, 1 August 2007, Pages 18-25
NeuroImage

Technical Note
Detecting hippocampal hypometabolism in Mild Cognitive Impairment using automatic voxel-based approaches

https://doi.org/10.1016/j.neuroimage.2007.04.048Get rights and content

Abstract

While the hippocampus is constantly reported as the site of earliest and highest structural alteration in Alzheimer's disease (AD), findings regarding the metabolic status of this region are rather heterogeneous. It has been proposed that only a time-consuming individual region-of-interest (ROI) approach would allow the detection of hypometabolism in this complex and small area. Our main goal with this study is to assess whether more automatic and clinically useful methods would be sensitive enough when considering other methodological confounds. From a single PET data set collected in 28 patients with amnestic Mild Cognitive Impairment (aMCI) and 19 controls, we assessed the effects of partial volume effect (PVE) correction, scaling (using vermis or global means), and analysis method (individual ROI versus more automatic template-based ROI or voxel-based approaches) on hippocampal hypometabolism detection in aMCI. PVE correction and scaling both showed a significant effect on group comparison, while the analysis method (individual versus template-based ROI) surprisingly did not. Hippocampal metabolic decrease was significant in all vermis-scaled conditions, and more so after PVE correction. Our findings highlight the crucial relevance of using reference-region-based (instead of global) scaling, and the higher sensitivity of PVE-corrected PET measures, to detect hippocampal hypometabolism in aMCI. They also show that hippocampal metabolic decline can be detected using template-based ROI as well as voxel-based methods. These findings have clinical relevance since they support the validity of more automatic and time-saving approaches to assess hippocampal metabolism changes in aMCI and early AD.

Introduction

Brain structural and functional alterations have been consistently demonstrated with MRI and FDG-PET in Alzheimer's disease (AD). Early changes have been observed at the stage of amnestic Mild Cognitive Impairment (aMCI), which represents the clinical group that best reflects the prodromal stage of AD (Petersen and Morris, 2005). In contrast to the growing evidence of early and marked structural alterations of the hippocampal region in AD (Nestor et al., 2004), findings regarding the metabolic status of this key area at the early stage of the disease have been somewhat divergent. Hippocampal hypometabolism has been reported in aMCI and mild AD in several PET studies using various methods (De Leon et al., 1997, Ouchi et al., 1998, DeSanti et al., 2001, Nestor et al., 2003, Anchisi et al., 2005, Mosconi et al., 2005, Mosconi et al., 2006) but not in others using either similar or different methodological approaches (Minoshima et al., 1994, Desgranges et al., 1998, Ibanez et al., 1998, Ishii et al., 1998, De Leon et al., 2001, Alexander et al., 2002, Herholz et al., 2002, Nestor et al., 2003, Mosconi et al., 2005, Mosconi et al., 2006, Kawachi et al., 2006), even though the finding of significant correlation between relative hippocampal metabolism and severity of episodic memory impairment suggested sensitivity was not the issue. The question of hippocampal metabolic preservation or alteration in early AD is crucial to support clinical diagnosis, to better understand the pathological mechanisms underlying this neurodegenerative disease and to guide pharmacological research.

The hippocampus is a small and complex structure particularly sensitive to methodological issues, which could thus have confounded previous PET results. Firstly, partial volume effect (PVE) corresponds to a limitation inherent to the camera and results in the contamination of metabolic values by those of neighboring voxels. This effect, which is exacerbated by atrophic process, could lead to hypometabolism overestimation. However, the effect of PVE correction on PET findings is not clear since one previous study reported significant hippocampal hypometabolism in aMCI only before, but not after, correcting for PVE (De Leon et al., 2001), while another study found a significant decrease in both conditions (Mosconi et al., 2005).

Secondly, in order to reduce inter-subject variability in global brain PET measures (which is known to affect also healthy subjects), PET data are scaled using either the global mean value of the whole brain, or the mean value of a known preserved cerebral region in AD (such as the pons or the cerebellar vermis). While the latter adequately reflects inter-individual variability, the former also reflects the pathological effect, i.e. a global decrease due to regional decreases and would therefore lead to an underestimation of regional changes (Buchert et al., 2005). However, both a presence and a lack of hippocampal hypometabolism were found after reference-region-based scaling (Herholz et al., 2002, Mosconi et al., 2006).

Finally, previous PET studies either used a region of interest (ROI) approach, which consists in manually delimiting the outline of a priori determined cerebral regions or a whole-brain voxel-based Statistical Parametric Mapping (SPM) method. Hippocampal hypometabolism has mostly been reported using the ROI approach but not with SPM. In addition, several studies using both methods on the same data reported a lower sensitivity of SPM compared to the ROI approach (De Leon et al., 2001, Nestor et al., 2003, Mosconi et al., 2005), which was thought to arise from the spatial resolution degradation required in the SPM procedure. However, the other abovementioned methodological confounds may also be involved. Moreover, previous studies used the SPM approach without a priori hypothesis upon the hippocampus, so that statistical thresholds were much more stringent than those used in ROI studies due to correction for multiple comparisons. For both methods to be compared, a similar hypothesis-driven approach (and thus comparable thresholds) should be used. Since the ROI method is particularly time-consuming and observer-dependant, it seems of particular clinical relevance to assess whether a more automatic method could prove to be sensitive enough and just as accurate to detect hippocampal dysfunction even in patients with aMCI.

Our main goal with this study is thus to determine the least demanding methodological approach allowing the accurate measurement of hippocampal metabolic changes in aMCI, as a more conservative pre-AD situation, by systematically assessing the potential confounding effects of PVE correction, scaling procedure and method of analysis.

Section snippets

Participants

Subjects' characteristics are summarized in Table 1. A total of 47 subjects were studied, including 28 patients with aMCI and 19 healthy elderly. aMCI patients were prospectively recruited through a memory clinic, which they all attended for a complaint of memory impairment. Following medical, neurological, neuropsychological, and neuroradiological investigations, they were selected according to previously described stringent criteria for aMCI (Chételat et al., 2005), i.e. isolated episodic

Results

The general ANCOVA (Table 2 and Fig. 3) showed, firstly, a significant main effect of Group only (aMCI < controls; Fig. 3a). Secondly, both PVE and scaling factors interacted with Group. Post hoc LSD tests revealed that PVE-corrected hippocampal metabolic values were higher than non-corrected values in controls (p = 0.007) but not in aMCI (p = 0.42; Fig. 3b) and that global scaling led to significantly lower hippocampal values than vermis scaling in both controls (p = 10 12) and aMCI (p = 10 8), the

Discussion

First, this study provides insights into previous literature discrepancies regarding hippocampal hypometabolism in prodromal AD, by clearly describing the influence of the main methodological confounds and highlighting their respective relevance when assessing this issue. Our data show that both PVE correction and scaling method should be considered, as suggested by their significant interaction with the Group factor. In contrast, the analysis methods (i.e. individual versus template-based ROI)

Acknowledgments

We acknowledge the financial support of the INSERM, PHRC (Ministère de la Santé) and Association France Alzheimer for this project. We thank C. Lalevée and A. Pélerin for help with neuropsychological assessments, B. Dupuy and D. Hannequin for their contribution to the recruitment of patients, M.H. Noël, M.C. Onfroy, D. Luet, O. Tirel and L. Barré for help with neuroimaging data acquisition, G. Kalpouzos for her help in healthy controls examination, and the individuals who participated in this

References (29)

  • G. Chételat et al.

    FDG-PET measurement is more accurate than neuropsychological assessments to predict global cognitive deterioration in patients with mild cognitive impairment

    Neurocase

    (2005)
  • M.J. De Leon et al.

    Cortisol reduces hippocampal glucose metabolism in normal elderly, but not in Alzheimer's disease

    J. Clin. Endocrinol. Metab.

    (1997)
  • M.J. De Leon et al.

    Prediction of cognitive decline in normal elderly subjects with 2-[18F]fluoro-2-deoxy-d-glucose positron-emission tomography (FDG-PET)

    Proc. Natl. Acad. Sci.

    (2001)
  • B. Desgranges et al.

    The neural substrates of memory systems impairment in Alzheimer's disease. A PET study of resting brain glucose utilization

    Brain

    (1998)
  • Cited by (77)

    • Effects of orientation change during environmental learning on age-related difference in spatial memory

      2019, Behavioural Brain Research
      Citation Excerpt :

      This integrative function of the MTL may be susceptible to age-related decline, leading to formation of less integrated and thus more distorted memories of large-scale environments. In addition, this decline may be exacerbated by deteriorated input into the MTL (i.e., inadequately performed conversion of spatial information between reference frames) due to metabolic alteration of the posterior cingulate region in the aged brain [56,57]. To refine and support these inferences, a neuroimaging study should be conducted in the future for directly contrasting activation in the MTL and posterior cingulate region while older adults learn large-scale environments with fixed and dynamic orientations.

    • PET/MRI Hybrid Systems

      2018, Seminars in Nuclear Medicine
    View all citing articles on Scopus
    View full text