In fMRI study, brain regions were found to be associated with verbal fluency tasks primarily located in the left hemisphere, including inferior and middle frontal gyrus, caudate, fusiform gyrus, and superior parietal lobe in healthy subjects (Birn et al., 2010). In addition, lesion studies have shown that the phonemic and semantic fluency tasks rely on different functional networks (J. V. Baldo et al., 2006; Henry &
Crawford, 2004; Troyer, Moscovitch, Winocur, Alexander, et al., 1998).
Specifically, temporal lobe lesions associate with semantic fluency while frontal lobe lesions tend to affect phonemic fluency more than semantic fluency. Recently, Birn et al. (2010) demonstrated different magnitude of activation in brain regions for contrasting two verbal fluency tasks respectively. For phonemic fluency, larger activation in left precentral and inferior frontal gyrus, and bilateral ventral occipito-temporal and superior parietal cortices were observed. For semantic fluency, greater activation in left middle frontal gyrus and bilateral occipital cortex, and fusiform gyrus were observed. In addition, anterior temporal lobes (ATLs) have been regarded as semantic hub which processing information from different motor and sensory areas to form amodal semantic representations (Patterson et al., 2015;
Patterson et al., 2007; Rogers et al., 2006). Overall, brain regions including frontal,
temporal, and parietal lobes are frequently associated with semantic memory (Binder et al., 2009).
Recent research has demonstrated the relationship between cerebral volumes and the performance of one form of category fluency task called the Issacs Set Test (IST) across time intervals (i.e. interval I: 0-15s；interval II: 16-30s) in elderly adults (Catheline et al., 2015). The results showed that lower score of IST was associated with smaller volumes in bilateral inferior frontal gyrus and right thalamus at interval I, while lower score at interval II was related to smaller volumes in left anterior
hippocampus and inferior parietal gyrus, reflecting a transition from frontal to
temopro-parietal regions across time when the word generation became more difficult.
Catheline et al. (2015) suggested that automatic speech production is dependent on executive functioning network integrity at the first 15se of the IST while controlled speech production at the later 15s is dependent on memory network integrity, including left hippocampus, which supported the notion that the temporal lobe structures underlie the proper functioning of semantic fluency task over time (Catheline et al., 2015). However, the dissociation between frontal regions and temporo-parietal regions for the early versus late phases in the IST (Catheline et al., 2015) has not been supported by other research yet, and was different to the notion that it was difficult to disentangle the language related processes from the executive
processes in a task which especially requiring rapid alternation of cognitive operations (Duffau et al., 2014) in such a short period of time. Therefore, we still inclined to divide the time into two 30-s intervals based on previous studies in various
populations, including young adults (Raboutet et al., 2010), cognitively normal older adults (Lee et al., 2015; Sauzéon et al., 2010), MCI (Weakley et al., 2013), and AD (Weakley & Schmitter-Edgecombe, 2014), in order to provide the utility of brain regions associated with category fluency tasks in these two time intervals. In addition to avoid the short time interval (i.e., 0-15s) might be influenced by the delayed first word production due to the deterioration of processing speed at the first 5s based on previous finding in normal aging adults (Lee et al., 2015).
With respect to the brain structures associate with semantic fluency in AD and MCI, some researchers have shown that semantic fluency deficits in AD are related to the reduction of gray matter (GM) volume in bilateral medial temporal lobes,
including bilateral middle temporal gyrus (Serra et al., 2010) and left fronto-temporal regions. (Venneri et al., 2008).
There also have been studies investigated the relationship between white matter (WM) integrity in AD and MCI, however, the results remain controversial. Some diffusion tensor imaging (DTI) studies found that semantic fluency performance was associated with fractional anisotropy (FA) value in the genu and splenium part of
corpus callosum, and frontal periventricular WM (T. F. Chen et al., 2009; Kavcic et al., 2008). In addition, T. F. Chen et al. (2009) demonstrated that frontal
periventricular WM was associated with the category fluency task in aMCI while Kavcic et al. (2008) found posterior callosal FA was associated with the performance of animal fluency in early AD. For the white matter changes in AD and MCI groups, Serra et al. (2010) found decreased FA values in anterior part of right anterior
thalamic radiation in aMCI; decreased FA values in splenium of corpus callosum (CC), right fornix, cingulum, and posterior thalamic radiation, and bilateral thalamic radiation FA in AD compared with controls. To compare with MCI, the authors found FA value was reduced in the left thalamic radiations and splenium of CC in AD.
Recently, Rodriguez-Aranda et al. (2016) investigated white matter tracts in early AD and found that reduced FA values in CC, forcep minor, bilateral inferior fronto-occipital fasciculus (IFOF) and superior longitudinal fasciculus (SLF), and the tracts were associated with the accurate scores of semantic fluency. The results showed a significant association between word latencies and language-related tracts known for semantic retrieval, such as bilateral uncinate fasciculus (UF), IFOF, forcep minor, and a majority of CC. This is consistent with previous researches show that IFOF and UF are subserving semantic processing in healthy older adults (de Zubicaray et al., 2011) and patients with brain lesion (Han et al., 2013). In addition, a majority of
studies have demonstrated that the left-sided IFOF and UF were involved in the semantic processing (Almairac et al., 2015; de Zubicaray et al., 2011; Han et al., 2013). Moreover, Jokinen et al. (2007) found that semantic verbal fluency was associated with the CC as a whole, and the isthmus subregion of CC was particularly important. Hippocampal gray matter (GM) atrophy has been regarded as a
vulnerable and early affected biomarker in aMCI (Devanand et al., 2007) or even in the premild MCI stage (Pre-MCI, i.e., a stage between elderly without cognitive impairment and aMCI) (Duara et al., 2011). It could also provide predictive values for detecting the progression to AD in three years (Whitwell et al., 2007).
There has been evidence showing that post-mortem AD patient also present white matter pathology, such as demyelination, axonal loss, and death of
oligodendrocyte microglial cells (Gouw et al., 2008). However, whether of the changes in white matter microstructure is independent of or secondary to gray matter pathology in MCI remains debatable. Recently, there is evidence supports the hypothesis that white matter degradation is independent of gray matter changes in prodromal stage of AD. For example, Y. L. Chang et al. (2015) demonstrated that multiple domain aMCI compared with healthy controls had significantly smaller GFA values in left inferior cingulum while there was no difference on gray matter integrity in bilateral hippocampi or cingulate regions. In addition, there have been studies
demonstrated white matter degradation without gray matter volume loss in subjective memory impairment, MCI (Selnes et al., 2012), and the group of healthy control later converted to aMCI in two years (Chang et al., 2016; Zhuang et al., 2012; Zhuang et al., 2013). Furthermore, AD animal model demonstrated that white matter damages (e.g., axonal blockage) interferes axonal transport preceding evident amyloid and other pathology occurs at least one year (Stokin et al., 2005). The cumulating evidence suggests white matter degeneration not only plays a role in the pathogenesis of cognitive decline but could happen before gray matter loss in the early stage of AD.
In order to discover the detail changes of white matter microstructures of related bundles of category fluency tasks in MCI, we adopted the tractography-based automatic analysis (TBAA) method to perform further investigation (Chen, 2015).
The TBAA method combines two type of information, first establishing a high-quality diffusion spectrum imaging (DSI) template and then reconstructing coordinates of the cerebral white matter tracts on the DSI template. Through the method, a
transformation map between DSI template and individual subject’s DSI data was created, and the microstructural properties (e.g., generalized FA, GFA) of a specific tract can be extracted from the subject’s data and analyzed in a voxel-by-voxel manner.