Although episodic memory deficits are the hallmark feature of Mild Cognitive Impairment (MCI), deficits in other cognitive domains are present in a large number of patients. In particular, mild anomia (Storandt et al., 2002
; Blackwell et al., 2004
) and reductions in semantic fluency [for a review, see (Taler and Phillips, 2008
)] develop in many patients with MCI, suggesting impaired lexical-semantic processing. In addition, a subset of patients with MCI show significant executive dysfunction, characterized by impaired working memory, inhibition, set-shifting, and phonemic fluency (Belleville et al., 2007
; Chang et al., 2009
). What has not been established is whether these domain-specific cognitive deficits in MCI are secondary to global brain atrophy versus progressive atrophy within specific neocortical regions. Medial temporal lobe atrophy is prominent in patients with Alzheimer’s disease (AD), but there is increasing evidence that atrophy is widespread even in preclinical AD (Fennema-Notestine et al., 2009
; McEvoy et al., 2009
). Therefore, delineating the MRI correlates of domain-specific cognitive decline could lead to an improved understanding of the neural basis of cognitive impairment in MCI.
There is an emerging literature describing baseline structural MRI correlates of cognitive impairment in MCI, AD, and mixed patient samples (Galton et al., 2001
; Grossman et al., 2004
; van der Flier et al., 2005
; Apostolova et al., 2008
). The most reliable and well-documented finding is an association between impaired verbal memory and medial temporal lobe atrophy that is particularly robust for hippocampal and entorhinal regions [see (Ries et al., 2008
) for a review]. Hippocampal and entorhinal atrophy have been shown to predict conversion to AD (Jack et al., 1999
; Jack et al., 2000
; Killiany et al., 2002
; Jack et al., 2005
; McEvoy et al., 2009
), as well as memory decline in MCI and AD (Mungas et al., 2001
; Cardenas et al., 2009
). Therefore, this relationship is the most frequently studied and the medial temporal lobes are the most common targets for region of interest MRI analyses in MCI.
Studies of lexical-semantic processing are fewer in number, but there is evidence linking impaired naming and semantic fluency to atrophy within a number of neocortical regions. Impaired visual naming has been linked to medial temporal lobe atrophy in a sample of healthy elderly, MCI, and AD (van der Flier et al., 2005
). In a study of patients with AD and MCI who later converted to AD, impaired visual naming and semantic fluency were associated with left parietal, and bilateral frontal, temporal lobe, and anterior cingulate atrophy (Apostolova et al., 2008
). Although there was considerable overlap in the regional atrophy associated with impairment on each task, reduced semantic fluency showed a stronger correlation with left inferior parietal and supramarginal atrophy, whereas reduced naming showed a stronger correlation with atrophy in left inferior temporal cortex. These studies provide evidence that language impairments in MCI and/or AD are linked to atrophy within a number of perisylvian regions, but that naming and semantic fluency have partially unique neuroanatomical substrates.
Structural MRI studies of executive functioning in MCI are scarce, but there are data suggesting that atrophy within frontal, cingulate, and temporal lobe regions contribute to executive dysfunction. In particular, dorsolateral and medial frontal lobe volume loss has been associated with poorer composite scores derived from measures of fluency, set-shifting, and response inhibition in MCI (Cardenas et al., 2009
). Neocortical thinning in dorsolateral frontal, posterior cingulate, and lateral temporal lobe regions has also been associated with impaired set-shifting and working memory performance (Chang et al., 2009
), and decreases in left dorsolateral and medial gray matter concentration has been detected in patients with a dysexecutive subtype of MCI (Pa et al., 2009
). These data support previous literature implicating dorsolateral prefrontal regions in executive functioning, but also suggest that executive dysfunction in MCI may be more complex, relying on the integrity of a number of other frontal and posterior cortical regions.
To date, only a handful of studies have employed longitudinal MRI for understanding the relationship between atrophy rates and cognitive decline in MCI. Whole brain atrophy rates have been associated with decline on measures of global cognition (Evans et al.; Sluimer et al., 2008
), verbal memory (Jack et al., 2004
), and set-shifting performance (Evans et al.). Total cortical gray matter and hippocampal atrophy has been linked to memory and executive functioning decline in MCI (Mungas et al., 2005
), and temporal lobe atrophy rates have been linked to both verbal memory decline and overall disease progression in MCI (Leow et al., 2009
These studies provide compelling evidence for a relationship between atrophy rates and cognitive decline in MCI. However, most existing studies have relied on baseline imaging or have evaluated longitudinal changes in whole brain atrophy or a very limited number of regions, providing a snapshot into critical structure-function relationships. Furthermore, many studies have included mixed MCI/AD/healthy control samples, precluding an analysis of whether the structure-function relationships are general in nature or unique to diagnosis. Therefore, the degree to which regional neocortical atrophy rates are related to domain-specific cognitive decline in MCI has not been fully evaluated.
In this study, we investigate the relationship between regional neocortical atrophy rates and domain-specific cognitive decline in a large, well-characterized group of patients with MCI. We evaluate whether atrophy rates obtained over a two-year period are related to memory, language, and executive function decline over the same time interval in MCI, and whether the atrophy patterns associated with decline in each cognitive domain are spatially unique from the pattern associated with increasing disease severity. Based on the existing literature, we predicted that left medial temporal lobe atrophy would be associated with verbal memory decline, whereas dorsolateral frontal lobe atrophy would be most related to executive functioning decline in MCI. We hypothesized that atrophy rates within left perisylvian regions would be associated with naming and semantic fluency decline in patients with MCI, but that left temporal lobe atrophy would contribute to naming decline, whereas left temporoparietal atrophy would contribute to semantic fluency decline.