Regular articleLongitudinal imaging pattern analysis (SPARE-CD index) detects early structural and functional changes before cognitive decline in healthy older adults
Introduction
An estimated 22–45% of older people in locations worldwide have cognitive impairment without dementia (Hanninen et al., 1996, Plassman et al., 2008, Scafato et al., 2010). Although any degree of cognitive impairment can negatively impact quality of life (Missotten et al., 2008, Missotten et al., 2009), increasing impairment often signifies the progression of life-altering illnesses such as Alzheimer's disease (AD). Biomarkers of early brain changes associated with AD are assuming increasing importance in newly revised criteria for research and clinical diagnoses (Albert et al., 2011, McKhann et al., 2011, Sperling et al., 2011). Because many intervention strategies, such as cognitive training (Mowszowski et al., 2010), physical exercise (Baker et al., 2010, Yaffe et al., 2001), and nutritional modifications (Naismith et al., 2009), may be more useful at earlier stages, it is important to detect cognitive decline early.
Given that it is generally believed that cognitive changes occur after structural and functional brain changes (Jack et al., 2009, Jack et al., 2010, Petersen and Jack, 2009), detecting changes in structural magnetic resonance imaging (MRI) and functional positron emission tomography (PET) cerebral blood flow (CBF) measures may be useful for early detection. However, the spatiotemporal characteristics of brain change in elderly individuals are widespread, complex, difficult to quantify, and difficult to detect on an individual basis with sufficient sensitivity and specificity. High-dimensional pattern analysis and classification methods may be useful for early identification of brain changes on an individual level, as they have been shown to capture these complex characteristics and produce quantitative metrics (Davatzikos et al., 2008, Davatzikos et al., 2009, Duchesne et al., 2008, Kloppel et al., 2008, Koutsouleris et al., 2009, Misra et al., 2009, Vemuri et al., 2008).
In this article, we aim to determine whether structural and functional patterns are detectable and quantifiable before cognitive decline in apparently healthy individuals by using a high-dimensional pattern analysis technique along with longitudinal analysis. We focus on cognitive decline measured using immediate free recall score from the California Verbal Learning Test (CVLT) (Delis et al., 1987), as change in immediate verbal recall is among the earliest cognitive changes detected during the preclinical phase of Alzheimer's disease (Grober et al., 2008). We further describe later in the text our choice of high-dimensional pattern analysis, along with a predominant model of cognitive decline.
Recent studies have shown the potential of high-dimensional pattern analysis and classification methods as a means to quantify and summarize imaging patterns (Davatzikos et al., 2008, Davatzikos et al., 2009, Kloppel et al., 2008, Koutsouleris et al., 2009, Misra et al., 2009, Vemuri et al., 2008). These methods also provide diagnostic and prognostic indicators for individuals, rather than groups, which is essential for clinical application. Pattern analysis and classification techniques use data labeled into classes, typically the presence and absence of a disease or condition, to derive a score representing the similarity of a given image to the image pattern representative of that condition. Such work has been useful for distinguishing, for example, normal subjects from those with AD (Davatzikos et al., 2009). However, characterizing patterns emerging at much earlier stages of cognitive decline before clinical impairment are ultimately going to be more important, as intervention at these stages before irreversible damage is likely to be more effective. To fill this need, we studied aging individuals with cognitive decline compared with those without cognitive decline from the Baltimore Longitudinal Study of Aging (BLSA). The BLSA provides imaging and cognitive data from older adults over long follow-up periods, allowing quantitative characterization of structural and functional data relative to cognitive decline.
A predominant hypothesis (Jack et al., 2009, Jack et al., 2010, Petersen and Jack, 2009) is that the progression of memory decline in AD occurs in the brain in the following order: 1) beta-amyloid protein deposits build up in the brain, causing plaques to develop, which may be an initial step in the chain of events leading to neuronal dysfunction and ultimately loss; 2) brain function decreases; 3) atrophy occurs, changing the structural properties of the brain; 4) cognitive ability declines, leading to declines in memory and executive function; 5) ultimately this leads to cognitive impairment across a range of cognitive domains. This progression is schematically shown in Fig. 1 (resembling a similar figure in [Jack et al., 2010]). In this study, we investigate further the temporal relationship of structural, functional, and cognitive decline.
Based on this model of cognitive decline, along with recent evidence that high-dimensional pattern classification can produce predictive and early diagnostic imaging-based biomarkers, we hypothesize that structural and functional change can be detected before cognitive test score decline in healthy older individuals. We also aim to determine a quantifiable score that characterizes patterns representative of cognitive decline during normal aging. Such a characterization could lead to identification of individuals at high risk for development of AD at very early stages of disease progression.
Section snippets
Methods
We use high-dimensional pattern classification techniques, combined with indicators of the dynamic changes of structure and function, to determine the temporal sequence of structural and functional brain changes and cognitive decline.
Change in brain structure and function precedes memory decline
We observed the temporal sequence of structural, functional, and cognitive decline. Having calculated the points of decline for all subjects with more than five time points, we computed the difference for each subject between the SPARE-CD point of decline and the CVLT score point of decline, visually represented in Fig. 2. MRI SPARE-CD scores declined on average 2.8 (standard deviation [SD] 2.5) years earlier than CVLT scores. [15O] PET-CBF SPARE-CD scores declined on average 2.3 (SD 2.6) years
Discussion
In this work, we study the functional and structural progression of changes related to cognitive decline in apparently healthy individuals, to identify very early imaging-based biomarkers for cognitive decline. We tested the hypothesis that spatial patterns of brain atrophy and change in cerebral blood flow, summarized by high-dimensional pattern classification and the SPARE-CD index, precede cognitive decline in a cohort of cognitively healthy older adults. We have shown that 1) there are
Conclusions
In this study, we investigated longitudinal progression of imaging characteristics of early cognitive decline during normal aging, well before clinically measurable changes occur, by leveraging on high-dimensional imaging pattern classification methods and imaging data from the BLSA. We produced an individualized quantitative value (SPARE-CD) representing functional ([15O] PET-CBF) and/or structural (MRI) change associated with cognitive (CVLT) decline, and in addition to finding that SPARE-CD
Disclosure statement
No conflicts of interest have been reported by the authors. All studies were approved by the institutional review boards of all institutions involved.
Acknowledgements
This work was supported by the National Institutes of Health (grant numbers R01-AG-14971, N01-AG-3-2124); and the Intramural Research Program of the National Institutes of Health, National Institute on Aging. The authors would like to thank Yang An at NIH for producing the CVLT score slopes using mixed-effects linear regression, Bennett Landman at Vanderbilt for providing a script for normalization of [15O] PET-CBF data, and Stathis Kanterakis and Drew Parker at Penn for assistance with
References (63)
- et al.
The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease
Alzheimers Dement
(2011) - et al.
Voxel-based morphometry: the methods
Neuroimage
(2000) - et al.
I. Longitudinal changes in aging brain function
Neurobiol. Aging
(2008) AFNI: software for analysis and visualization of functional magnetic resonance neuroimages
Comput. Biomed. Res. Int. J
(1996)- et al.
Voxel-based morphometry using the RAVENS maps: methods and validation using simulated longitudinal atrophy
Neuroimage
(2001) - et al.
Individual patient diagnosis of AD and FTD via high-dimensional pattern classification of MRI
Neuroimage
(2008) - et al.
Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade
Lancet Neurol
(2010) - et al.
A 3D atlas of the human brain
Neuroimage
(1998) - et al.
The diagnosis of dementia due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease
Alzheimers Dement
(2011) - et al.
Baseline and longitudinal patterns of brain atrophy in MCI patients, and their use in prediction of short-term conversion to AD: results from ADNI
Neuroimage
(2009)
Very high-resolution morphometry using mass-preserving deformations and HAMMER elastic registration
Neuroimage
Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease
Alzheimers Dement
Alzheimer's disease diagnosis in individual subjects using structural MR images: validation studies
Neuroimage
High-dimensional pattern regression using machine learning: from medical images to continuous clinical variables
Neuroimage
Improvement of brain segmentation accuracy by optimizing non-uniformity correction using N3
Neuroimage
Linear regression with spatial constraint to generate parametric images of ligand-receptor dynamic PET studies with a simplified reference tissue model
Neuroimage
Using a reference tissue model with spatial constraint to quantify [11C]Pittsburgh compound B PET for early diagnosis of Alzheimer's disease
Neuroimage
Effects of aerobic exercise on mild cognitive impairment: a controlled trial
Arch. Neurol
Stability of default-mode network activity in the aging brain
Brain Imaging Behav
Hypercortisolism in alcohol dependence and its relation to hippocampal volume loss
J. Stud. Alcohol
Hippocampus volume loss due to chronic heavy drinking
Alcohol. Clin. Exp. Res
The association between quantitative measures of dementia and of senile change in cerebral grey matter of elderly subjects
Br. J. Psychiatry
Reduced resting-state brain activity in the “default network” in normal aging
Cereb. Cortex
Longitudinal progression of Alzheimer's-like patterns of atrophy in normal older adults: the SPARE-AD index
Brain
California Verbal Learning Test—Research Edition
Longitudinal pattern of regional brain volume change differentiates normal aging from MCI
Neurology
MRI-based automated computer classification of probable AD versus normal controls
IEEE Trans. Med. Imaging
The medial temporal lobe and recognition memory
Annu. Rev. Neurosci
COMPARE: classification of morphological patterns using adaptive regional elements
IEEE Trans. Med. Imaging
An image-processing system for qualitative and quantitative volumetric analysis of brain images
J. Comput. Assist. Tomogr
Age-related changes in brain activity across the adult lifespan
J. Cogn. Neurosci
Cited by (29)
Effects of trihexyphenidyl on prefrontal executive function and spontaneous neural activity in patients with tremor-dominant Parkinson's disease: An fNIRS study
2022, Parkinsonism and Related DisordersCitation Excerpt :The commonly recommended daily dose of THP for PD is 2–8 mg [27]; the average daily THP dose was relatively low in this study, at 3.1 mg. In individuals with cognitive decline, the order of decline in brain markers tends to manifest first as a decline in functional neuroimaging, then as brain structural changes, and finally as a decline in cognitive ability [28]. Thus, we speculate that the THP effect on cognitive function is dose and time dependent, and the low and non-prolonged doses in our study mean that the brain activation of patients with tdPD is affected first.
A survey on machine and statistical learning for longitudinal analysis of neuroimaging data in Alzheimer's disease
2020, Computer Methods and Programs in BiomedicineCitation Excerpt :This is an important challenge that needs to be addressed, and it is especially critical in progression modelling and temporal prediction of the disease. Some of the reference variables used in the literature are time from baseline acquisition [47,114], age of the subject [142], normalized age (where age variation has been removed [127]), cognitive scores [38,116], data-driven progression scores [45,82,83,87,129] or any other type of data-driven temporal alignment [123]. Methods that account for alignment provide a more complete interpretation of the disease progression over time, as they allow studying between-subjects differences and assessing the evolution of a given patient with respect to that of the general population.
Evaluating Alzheimer's disease biomarkers as mediators of age-related cognitive decline
2017, Neurobiology of AgingIntegration and relative value of biomarkers for prediction of MCI to AD progression: Spatial patterns of brain atrophy, cognitive scores, APOE genotype and CSF biomarkers
2014, NeuroImage: ClinicalCitation Excerpt :The predictive value of SPARE-AD in MCI individuals complements earlier studies that found similar AD-like patterns of brain atrophy being predictive of cognitive decline in cognitively normal older adults (Clark et al., 2012b; Dickerson and Wolk, 2012). Particularly relevant is our previous study (Clark et al., 2012b), since it used the exact same image analysis and SPARE-AD index. In that prospective longitudinal study of aging over an 8-year period, the rate of change of SPARE-AD was highly predictive of conversion from cognitively normal to MCI, with a cross-validated AUC of 0.89.
Memory decline shows stronger associations with estimated spatial patterns of amyloid deposition progression than total amyloid burden
2013, Neurobiology of AgingCitation Excerpt :Although cross-sectional studies of relationships between amyloid burden and cognition have yielded mixed results, higher amyloid burden has been associated with greater longitudinal memory decline in a number of studies (reviewed by Resnick and Sojkova, 2011). In prior work, we have demonstrated associations between rates of longitudinal change in verbal episodic memory performance and structural/functional changes in the cortex (Clark et al., 2012). Thus, in the present study, we aimed to use rates of change in California Verbal Learning Test (CVLT) scores to group participants for comparison of amyloid progression patterns.