Altered Topology in Information Processing of a Narrated Story in Older Adults with Mild Cognitive Impairment

Abstract

The ability to store, integrate, and manipulate information declines with aging. These changes occur earlier, faster, and to a greater degree as a result of neurodegeneration. One of the most common and early characteristics of cognitive decline is difficulty with comprehension of information. The neural mechanisms underlying this breakdown of information processing are poorly understood. Using functional MRI and natural stimuli (e.g., stories), we mapped the neural mechanisms by which the human brain accumulates and processes information with increasing duration and complexity in participants with amnestic mild cognitive impairment (aMCI) and healthy older adults. To explore the mechanisms of information processing, we measured the reliability of brain responses elicited by listening to different versions of a narrated story created by segmenting the story into words, sentences, and paragraphs and then scrambling the segments. Comparing healthy older adults and participants with aMCI revealed that in both groups, all types of stimuli similarly recruited primary auditory areas. However, prominent differences between groups were found at the level of processing long and complex stimuli. In healthy older adults, parietal and frontal regions demonstrated highly synchronized responses in both the paragraph and full story conditions, as has been previously reported in young adults. Participants with aMCI, however, exhibited a robust functional shift of long time scale processing to the pre- and post-central sulci. Our results suggest that participants with aMCI experienced a functional shift of higher order auditory information processing, possibly reflecting a functional response to concurrent or impending neuronal or synaptic loss. This observation might assist in understanding mechanisms of cognitive decline in aMCI.

Keywords

Aging functional MRI mild cognitive impairment processing time scales real-life stimulation

INTRODUCTION

Mild cognitive impairment (MCI) refers to a transitional state between normal aging and dementia [1, 2]. Amnestic MCI (aMCI), the most common subtype, is characterized by impairment in information and memory processing, and a high risk of conversion to dementia [3 –6] and Alzheimer’s disease (AD). Persons with aMCI show impaired performance in tests of episodic memory, especially in delayed recall [7, 8]. In addition, decline in language function has been commonly demonstrated in aMCI [7 –10]. These cognitive impairments are associated with brain neurodegenerative processes [11, 12], in particular of the hippocampus and medial temporal lobe structures [3 , 13–16].

Previous functional magnetic resonance imaging (fMRI) studies have explored differences in brain activation of individuals with aMCI using problem solving [17] or encoding and retrieval memory tasks (for review see [18]), with hyper- or hypoactivation observed in some medial temporal regions as compared to healthy controls. However, many fMRI studies to date have used relatively simple stimuli in highly controlled and artificial setups [19 –27]. The processing of information in real-life situations is not necessarily mimicked by these limited conditions. The ability to process a continuous flow of information is required for comprehension of most complex situations, as processing occurs at different time scales (e.g., words, sentences, and narratives) [28], and relies on accumulation of information over time. Some fMRI studies have explored more complex information processing in aMCI [29] but the stimuli used were short (lasting seconds) and did not map the neural responses unfolding over time (from seconds to many minutes).

Here we applied a completely different approach to identify possible changes in the mechanisms underlying information processing in aMCI. We investigated the neural responses to an ecological and complex stimulus, a narrated story, in participants with aMCI. Furthermore, we used inter-subject correlation (inter-SC), which is an advanced, fMRI-based analysis that measures the reliability of evoked-neural responses between participants, in order to identify differences in neural processing of complex stimuli between subject groups. Previous studies on young healthy participants suggested that a topographical hierarchy of brain areas is involved in processing of visual [30] and auditory [31] information over different time scales. Analogous with the concept of a spatial receptive field [32], a temporal receptive window (TRW) of a neural circuit has been defined as the length of time prior to a neural response in which sensory information may affect that response [30]. Depending on the capacity to accumulate sensory information over time, TRWs become progressively longer as one moves from low-level sensory to higher-level perception and cognition, and different brain areas are recruited accordingly, with low-level sensory information recruiting primary sensory cortex and higher level processing recruiting additional supplementary cortex in a hierarchical fashion. This topographical hierarchy was demonstrated using a narrated story and versions of the same story in which the words, sentences, and paragraphs of the full story were segmented and scrambled. Each scrambled version of the story contained coherent information only at the level of a unique language time scale (i.e., words, sentences, paragraphs). This paradigm enabled measuring neural responses to the different TRWs by varying degrees of temporal, and thus, contextual complexity. We used this approach to define and compare the topology of neural responses to TRWs in healthy older adults and participants with aMCI, in order to investigate the mechanisms underlying comprehension in aMCI. Specifically, we aimed at exploring the hierarchy of brain regions involved in information processing over different time scales in participants with aMCI in comparison to healthy older adults.

Comparing healthy older adults and participants with aMCI would allow us to explore whether patterns of information processing are different in normally aging brains as opposed to brains that are prone to or in the beginnings of degenerative disease. In addition we aimed to study whether any altered patterns of processing may have a functional significance, as indicated by an association with performance on neuropsychological tests of information processing and memory.

MATERIALS AND METHODS

Participants

Fifteen healthy older adults (4 males, mean age 67.3±6.20 y) and fourteen patients with aMCI (8 males, mean age 72.1±5.44 y) were studied. All participants were fluent Hebrew speakers. Healthy participants were recruited from the local community using accepted methods, mainly through ads posted at local community centers. Healthy older participants were fully independent in all instrumental activities of daily living, and reported no previous neurological or psychiatric symptoms or hearing impairments that would affect their participation in the study. They were first screened by a comprehensive neuropsychological assessment, which included standardized measures of memory, attention, language, and executive function (see a full description of the testing battery below). Subjects who scored either less than 25 on the Montreal Cognitive Assessment (MoCA), a standard screening test for mild cognitive dysfunction [33, 34], or more than 1.5 standard deviations below average on any of the cognitive tests were not considered healthy and thus were excluded from participation.

Participants with aMCI were recruited from the outpatient clinic of the Center for Memory and Attention Disorders at Tel-Aviv Sourasky Medical Center. They fulfilled the same inclusion and exclusion criteria as the healthy participants, but were diagnosed with aMCI by an expert neurologist according to published criteria [12].

The Human Studies Committee of Tel-Aviv Sourasky Medical Center approved the study. All participants provided written informed consent to participate in the study.

MRI acquisition

MRI scanning was performed at Tel-Aviv Soura-sky Medical Center with a 3T head-only MRI scanner (GE Signa EXCITE, Milwaukee, WI, USA), using an eight-channel head coil. Blood oxygenation level dependent (BOLD) functional MRI was acquired with T2^*-weighted imaging using the following parameters: repetition time (TR) = 1500 ms, 492 TRs in each run; echo time (TE) = 30 ms; flip angle (FA) = 75°; field of view (FOV) = 22×22 cm²; matrix size = 64×64; 27 slices of 3 mm thickness, 1 mm gap. The slices were positioned nearly horizontal to cover the entire temporal lobe and the parts of the frontal lobe that are involved in hearing and language processing, as well as nearly all of the occipital and parietal lobes. High-resolution, anatomical, T1-weighted, fast-spoiled gradient-echo images were acquired using the following parameters: FOV = 256 mm; matrix = 256×256; TR = 9.2 ms; TE = 3.5 ms; axial slices of 1 mm thickness, no gap. These anatomical volumes were used for cortical segmentation and surface reconstruction.

To minimize head movements, participants’ heads were stabilized with foam padding. MRI-compatible headphones (OPTOACTIVE^tm) were used to considerably attenuate the scanner noise and to present the audio stimuli (Presentation^®, Neurobehavioral Systems).

Stimuli and experimental design

Stimuli for the experiment were generated from a 12-min, real-life story, told in Hebrew by a professional storyteller and recorded specially for the study. In the scanner, participants listened to the full story from beginning to end (forward condition), as well as to the story presented waveform-reversed in time (backward condition). Participants also listened to three other stimuli created by segmenting the full story into different elements of language structure and then scrambling them. Specifically, the story was segmented manually by identifying the end points of each word, sentence, and paragraph, thus defining three time scales: short (words), intermediate (sentences), and long (paragraphs and story; Fig. 1). In each of the conditions, the isolated language structures were then scrambled by a random shuffling of the order of the segments. Thus, all scrambled conditions were “reorderings” of the same forward condition, either as a scramble of individual words (“words scram” condition), individual sentences (“sentences scram” condition), or paragraphs (“parag scram” condition). Each scrambled condition served to isolate the contextual information available for comprehension as a function of that specific language structure, thus yielding the three different time scales for information processing. By comparing neural responses evoked by these different conditions, we were able to characterize the informational capacity of brain regions involved in processing of information over different time scales. As described above, three types of scrambled conditions were obtained: (1) Short stimulus (“words scram”) included 1104 words (a word lasted on average 0.66±0.33 s); (2) Intermediate stimulus (“sentences scram”) included 123 sentences (a sentence lasted for 5.94±2.95 s, and (3) Long stimulus (“paragraphs scram”) included 25 paragraphs (a paragraph lasted for 29.24±5.55 s). The overall durations of the forward, backward, and the scrambled stimuli were identical. For example, the short scrambled condition (“words scram”) was the same overall length as the original story from which it was derived. This was true for all of the conditions tested. Three seconds of silence preceded and also followed each condition and were discarded from all analyses.

A typical session was comprised of 5 runs, each consisting of the presentation of 1 of the 5 different conditions. To minimize repetition effects, the presentation order of all conditions was randomized across participants. Attentive listening to the story and its scaled variants was confirmed by brief communication with participants between runs. In addition, participants were asked to freely describe the main characters of the plot at the end of the scanning session, outside the scanner, and were able to accurately recount the plot and story line.

Neuropsychological assessment

All participants underwent a comprehensive cognitive battery which included the MoCA test, as well as the following standardized tests of cognitive function. Tests of executive function and language consisted of the color version of the trail making test (CTT) [9 , 36], the verbal fluency test (semantic and phonemic) [35 , 37–39], and the digit span test [35, 39]. Memory tests included the Rey auditory verbal learning test (RAVLT) [9, 35], the Rey-Osterrieth complex figure [35], and the Wechsler logical memory scale [9, 35]. All tests were scored and standardized using appropriate, age-based norms, except the MoCA, where age-based norms are not available.

Data analysis

fMRI

Preprocessing: fMRI data were analyzed with the BrainVoyager QX software package (Brain Innovation, Maastricht, The Netherlands) and with in-house software written in MATLAB. Preprocessing of the functional scans included slice time and motion correction, linear trend removal, high-pass filtering (cut-off: 0.01 Hz), spatial smoothing using a Gaussian filter of 6 mm full-width at half-maximum value, and cropping of the first 40 TRs in each run to allow the hemodynamic responses to reach steady state. The cortical surface was reconstructed from the anatomical images using standard procedures implemented in the BrainVoyager software. Data analysis was performed separately for each subject. The two-dimensional functional images were co-registered with two-dimensional anatomical images using a 2-step, semi-automatic procedure. First, an initial alignment by Brain Voyager was performed. Next, a subsequent advanced manual alignment was applied. Finally, the data were incorporated into the three-dimensional data sets through trilinear interpolation. The complete functional dataset was transformed to a common 3D Talairach space [40] and projected on a reconstruction of the cortical surface. Runs in which head motions were greater than 2 mm were excluded from the analyses. Also, runs in which the signal was corrupted (e.g., contained spikes in the fMRI time series greater than a 5 standard deviation change in image intensity) were discarded.

Inter-subject correlation analysis: Data were analyzed using the inter-SC approach, exploring to what extent similar brain regions of different participants show synchronization of neural responses to natural stimuli. This synchronization was exhibited by significant correlation between each participant’s neural response and the other participants’ responses (inter-SC), indicating response reliability. Inter-SC maps were constructed voxel-by-voxel in Talairach space, separately for each condition (forward, backward, words scram, sentences scram, paragraphs scram) and for each group by comparing the fMRI response time courses across participants. First, the Pearson product-moment correlation was computed between a voxel’s fMRI time-course in one individual and the average of that voxel’s fMRI time-courses in the remaining participants. Next, the average correlation $R = \frac{1}{N} \sum_{j = 1}^{N} r_{j}$ was calculated at every voxel.

Temporal receptive window (TRW) maps: Following Lerner et al. [31, 41], we constructed a nested map of TRWs (Fig. 2), classifying the voxels according to the shortest temporal structure that evoked reliable responses. The map demonstrated a hierarchy of brain areas, where voxels at the base of the hierarchy responded reliably to all scrambling conditions, and areas at the highest level of the hierarchy responded reliably to only the forward condition. Specifically, red voxels responded reliably to all conditions; yellow voxels responded reliably to words scram, sentences scram, paragraphs scram and forward; green voxels responded reliably to sentences scram, paragraphs scram, and forward; blue voxels responded reliably to paragraphs scram and forward; and brown voxels responded reliably to only the forward condition. In detail, if correlations in a voxel were statistically significant for all conditions regardless of scrambling, then that voxel was labeled “backward” (Fig. 2, red), since the backward condition contains the shortest coherent temporal structure. Using the same process, if correlations in a voxel were significant for all conditions except backward then the voxel was labeled “words scram” (Fig. 2, yellow), which indicates a short TRW. A voxel was labeled “sentences scram” if correlations were significant for all conditions except backward and words (Fig. 2, green), indicating an intermediate TRW. A voxel was labeled “paragraphs scram” if correlations were significant in all maps except backward, “words scram”, and “sentences scram” (Fig. 2, blue), indicating a long TRW. Finally, voxels were labeled “forward” if correlations were significant for the full story only (Fig. 2, brown).

In a post-hoc analysis to further examine the effect of age on patterns of information processing, we defined two subgroups. The first subgroup included the 6 oldest healthy adults (67–78 y, mean 73±4.38) and the second subgroup included the 6 youngest participants with aMCI (62–72 y, mean 67.16±3.86). We then constructed TRW maps for each of the subgroups using the procedure described above.

Statistical analysis

Statistical significance of the inter-SC analyses was assessed using a phase-randomization procedure. Phase-randomization was performed by applying a fast Fourier transform to the signal, randomizing the phase of each Fourier component, and then inverting the Fourier transformation. Thus, the power spectrum was preserved but the correlation between any pair of such phase-randomized time-courses had an expected value of 0. Phase-randomized time courses were generated for every measured fMRI time course from every voxel in each participant. A correlation value was then computed (as detailed above) for every voxel. This process was repeated 5000 times to generate a null distribution of the correlation values, separately for each voxel. Statistical significance was assessed by comparing empirical correlation values (without phase randomization) to these null distributions. The Benjamini–Hochberg–Yekutieli false-discovery procedure, which controls the false discovery rate (FDR) under assumptions of dependence, was used to correct for multiple comparisons [42 –44]. Specifically, p-values were sorted in ascending order and the value was chosen as the p-value corresponding to the maximum such that $p_{k} < \frac{k}{N} q^{*}$ , where = 0.05 is the FDR threshold, and N is the total number of voxels.

The differences in neural response reliability between groups were assessed using a t-test. Specifically, a 1-tailed t-test was applied in voxels that were not rejected in the statistical procedure described above in at least one group.

Identifying regions of interest (ROIs) with different processing time scales

In order to further define the magnitude of the reliability of evoked-neural responses between participants, we identified ROIs with short, intermediate and long TRWs that had been previously obtained from an independent functional data set (healthy young participants in Lerner et al. [31]). The TRW of each region was defined according to the response reliability in that region across stimuli scrambled on the aforementioned different time scales. Namely, ROIs with a predicted TRW were selected: primary auditory cortex and adjacent areas (A1+), middle superior temporal gyrus (mSTG), posterior superior temporal gyrus (pSTG), temporo-parietal junction (TPJ), dorsolateral prefrontal cortex, and precuneus. In addition, the post-central sulcus (post-CS) ROI was defined anatomically.

Functional Connectivity Analysis (FCA)

Functional connectivity of a seed region was explored in the forward condition. The precuneus was selected as the seed ROI for the FCA analysis because it was the only region that exhibited reliable responses in the inter-SC analysis of the forward conditions in both healthy and aMCI groups. The precuneus ROI for this analysis was obtained from the responses of participants with aMCI since we aimed at defining a functional network that was specific for aMCI. Specifically, the seed precuneus ROI used for the FCA was a subpart of the precuneus ROI defined in the ROI analysis described above. To perform the FCA, seed time courses for the forward condition were extracted for each subject by averaging the time courses of all voxels in the ROI. The resulting individual ROI time courses were then entered into a multi-subject random-effects General Linear Model (RFX GLM) as regressors in a voxel-by-voxel correlation analysis (p < 0.005, FDR corrected). The final map reflected regions whose activity was correlated to the activity in the precuneus across participants within each group.

Structural analysis

To test whether functional decline in participants with aMCI was associated with structural changes, we performed volumetric measurement of the hippocampal and cortical regions in participants with aMCI. The volumetric analysis was performed on standard resolution 3D T1 weighted imaging axial images. Analysis was performed using the FreeSurfer v5.1 image analysis suite, an automated, well documented and freely available software (http://surfer.nmr.mgh.harvard.edu/) used for brain segmentation based on probabilistic atlas and intensity values. Briefly, the automated procedure included skull stripping, intensity normalization, Talairach transformation, tissue segmentation, and surface tessellation [45 –47]. The complete FreeSurfer analysis pipeline was performed with manual intervention and quality assurance of the data.

Neuropsychological assessment

Student’s t-tests were applied to establish between-group differences in age and neuropsychological test scores. For the RAVLT, the 1st, 5th, and 8th repetitions were chosen for the analysis as they represented baseline, learning and delayed memory, respectively. Statistical analyses were performed using SPSS software v20.

To test a relationship between memory function and reliability of brain responses in participants with aMCI, first, a Memory Index (MI) based on the average normalized scores in the tests of delayed recall of episodic memory was calculated for each individual. Specifically, the MI was calculated as the average of the following scores: 1) delayed recall of the RAVLT (8th repetition), 2) logical memory-delayed recall, and 3) Rey-Osterrieth complex figure-delayed recall score. These tests have been suggested to be most sensitive to the memory decline in individuals with MCI [48, 49]. Also, using an average score based on multiple delayed episodic memory tests has been shown to be more robust than a single test [48, 49]. Next, correlations between MI and inter-SC coefficients were calculated in two ROIs— the precuneus and the post-CS using Pearson product-moment correlation.

RESULTS

Neuropsychological results

Scores in the neuropsychological assessment are summarized in Table 1. The average MoCA score of participants with aMCI was 22.93±2.61, which was significantly lower than healthy older adults: 26.80±1.32 (p < 0.001). Significant group differences between healthy older adults and adults with aMCI were found on all memory tests, with the most significant differences observed in delayed recall (see Table 1). The MI was significantly lower in participants with aMCI as well. Semantic fluency scores were also significantly lower in the aMCI group (p = 0.002), which is consistent with the diagnosis of aMCI [9]. No other significant group differences were observed, again consistent with the aMCI diagnosis. Note that although the participants with aMCI were significantly older than the healthy cohort (p = 0.036), the scores from all cognitive tests other than the MoCA were standardized for age. All participants, regardless of their scores on the cognitive testing, could retell the story after the fMRI session.

Response reliability across brain regions for different time scales

A voxel-wise analysis showed a hierarchy of progressively longer TRWs (Fig. 2) in both groups. Although the hierarchy started similarly in A1+ for both groups, the topographic organization in higher-order regions was very different between groups. The topographic organization identified in healthy older adults (Fig. 2A) was similar to that observed in young participants [31]. Namely, response reliability in the primary auditory cortex (A1+) was essentially invariant to temporal scrambling, suggesting that this region has relatively short TRWs. Response reliability in higher brain areas depended on information accumulated over longer time scales, revealing a hierarchy of TRWs spanning short, intermediate, and long temporal scales, as observed in young subjects.

The reliability patterns in participants with aMCI and healthy older adults were very similar in the regions with short and intermediate TRWs. In both groups, primary auditory areas exhibited reliable responses to all conditions regardless of the level of scrambling (Fig. 2, red). Reliable responses for the “words scram” (Fig. 2, yellow, A1+) and “sentences scram” conditions (Fig. 2, green, mSTG) were also not significantly different in both groups. However, regions that demonstrated reliable responses for the longest time scales (i.e., “paragraphs scram” and forward conditions) in healthy older adults and participants with aMCI were prominently different. While reliable responses evoked by the “parag scram” (Fig. 2A, blue) and forward (Fig. 2A, brown) conditions in the healthy older adults were found in the pSTG, TPJ, precuneus, and frontal cortex, participants with aMCI exhibited reliable responses for long time scales in the pre- and post-CS (Fig. 2B, blue & brown). Voxel-wise, one-tailed, two-sample t-tests, applied to Pearson correlation coefficients (see Methods: inter-SC analysis) in the “parag scram” and forward conditions revealed statistically significant differences between healthy and aMCI groups (Fig. 3A, B). Specifically, in the areas along STG, TPJ, dorsolateral prefrontal cortex (DLPFC), inferior frontal sulcus (IFS), temporal pole, precuneus and mPFC (blue, Fig. 3), inter-SCs were more reliable for the healthy older adults than for participants with aMCI. In contrast, inter-SCs were more reliable for participants with aMCI than for healthy older adults in post-CS, CS, and left pre-CS (red, Fig. 3). Table 2 provides Talairach coordinates for the noted regions.

Since healthy older adults were significantly younger than participants with aMCI (p = 0.036), a post hoc analysis was applied to examine the effect of age on the results. To that end, we defined two subgroups. The first subgroup included the 6 oldest healthy adults and the second subgroup included 6 youngest participants with aMCI (see Methods for details). The average age of the subgroup of the oldest healthy older adults (73±4.38) was similar to the average age of the entire aMCI group (72.1±5.44). Similarly, the average age of the subgroup of younger participants with aMCI (67.16±3.86) was similar to the average age of the entire healthy older adults group (67.3±6.20). We then constructed TRW maps for each of the subgroups using the procedure described above. The results for these subgroups were similar to those observed for the groups overall, suggesting that age was not the cause for the differences in TRW hierarchy between participants with aMCI and healthy older adults (Supplementary Figure 1).

To further quantify the differences observed in the reliability of evoked responses in the two groups, we performed an ROI analysis of the regions identified in the voxel-wise inter-SC analysis. ROIs were defined on the independent data set obtained from the Lerner et al. (2011) study (for details see Methods). A clear topography of temporal structure, in which the reliability of responses to scrambled speech declined gradually, was observed with this analysis (Fig. 4). Primary auditory areas (A1+) with short TRWs and higher-order areas (mSTG, pSTG) with intermediate TRWs exhibited a similar level of inter-SC in both groups. However, areas with especially long TRWs, where the cortical activity at every moment depends on many seconds of preceding information (for example, TPJ and DLPFC in healthy adults) did not exhibit reliable responses in participants with aMCI. Moreover, individuals with aMCI demonstrated significantly reliable responses to the long time scale conditions in the pre- and post-CS. Of note, the precuneus was the only higher-order region evoking similar patterns of reliable responses for long time scales in both groups.

Functional connectivity

After observing these responses to long time scale conditions in the pre- and post-CS in participants with aMCI, we were interested in examining to what extent this effect can be explained by connectivity of these regions with other brain areas. In an attempt to answer this question, we performed functional connectivity analyses based on the ‘seed’ time course obtained during the forward condition. The precuneus was chosen as a seed region based on the results of the inter-SC analysis because it demonstrated reliable responses to the longest time scales in both groups. We then used these time courses as GLM predictors to compute a voxel-by-voxel fit (see Materials and Methods for details). Fit evaluation revealed an interesting effect. For healthy older adults, strong correlations (p < 0.001) between activity in the precuneus and in frontal lobe areas (e.g., DLPFC, Broca’s area, IFS), temporal lobe (e.g., STG, TPJ), and occipital lobe regions were found. However in participants with aMCI, connectivity of the precuneus was relatively limited. Strong correlations were observed with the DLPFC in areas adjacent to the CS and in the temporal lobe, but overall patterns of functional connectivity were less widespread than in healthy older adults. Figure 5 shows the one-tailed, t-test analysis between connectivity coefficients for older adults and participants with aMCI. Significant differences in the connectivity of the precuneus (Fig. 5A) to the TPJ were observed in healthy participants (blue in Fig. 5B) and to the post-CS and CS in participants with aMCI (red in Fig. 5B).

Relationship between memory function and reliability of brain responses

While the results of the inter-SC analysis show that the neural responses of participants with aMCI and healthy older adults to the processing of auditory information in long TRWs are different, the question still remains as to whether these differences are important to memory. To further explore this question, we examined the association between memory function and the reliability of brain responses in the aMCI group. We computed the correlation between Memory Index (MI) (average normalized scores in the tests of delayed recall, see Methods) and response reliability (as indexed by inter-SC coefficients for the forward condition) in two ROIs: the precuneus and post-CS. The precuneus was chosen because it was the only high order cognitive area that showed reliable responses in both groups. The post-CS was chosen as this was the area that revealed reliable responses in participants with aMCI in contrast to reliable responses observed in higher-order areas in both younger and older healthy adults. The inter-SC coefficients were plotted against MIs, and linear regression was performed (Fig. 6). Results of this analysis revealed a significant relationship in both tested ROIs. Specifically, higher MIs (better memory function) were associated with a higher correlation coefficient (R² = 0.4032, p = 0.008) in the precuneus, and lower MIs (worse memory function) were associated with a higher correlation coefficient (R² = 0.592, p < 0.001) in the post-CS.

Structural analysis

Volumetric measurement of cortical regions (cortical thickness) and intracranial volume of hippocampal regions revealed minor but significant differences between groups in the right and left inferior temporal gyrus (ITG, p = 0.006), the left lateral aspect of the STG (p = 0.032) and the right middle temporal gyrus (MTG, p = 0.030). Marginal differences were found in the left precuneus (p = 0.052). No differences were found in hippocampal volumes (Supplementary Table 1).

DISCUSSION

The current study has shown differences in information processing of a narrated story in older adults with aMCI as compared to healthy older adults. Our paradigm enabled an ecological observation of brain function while processing simple and complex real-life stimuli. Specifically, we measured neural responses to a narrated story and to temporally scrambled versions of that story in both groups. The results revealed a topographical hierarchy of neural responses in healthy older adults that was similar to the hierarchy found in younger adults [31]. In healthy older adults and participants with aMCI, we observed a similar pattern of reliable responses to short and intermediate temporal scales (from A1+ up to pSTG), suggesting that less complex temporal stimuli are being processed in a similar fashion in both groups. However, the typical hierarchy for processing complex auditory information was altered in aMCI; the synchronized responses for both the paragraphs scrambled and forward conditions were found in different regions for participants with aMCI as opposed to healthy subjects. Specifically, participants with aMCI exhibited reliable responses in areas that are not typically involved in the processing of complex auditory stimuli (the pre- and post-CS), but did not show reliable responses in higher order cognitive areas that are typically associated with this task (namely the pSTG, TPJ, and frontal cortex).

Possible mechanisms underlying changes in information processing patterns in aMCI

One possible mechanism for this shift in information processing of complex auditory stimuli in aMCI may be a compensation effect, where less affected brain areas may be taking over processing functions for other areas in various stages of neurodegeneration. In general, the pre- and post-CS are areas that are not typically observed as hotspots of atrophy or degeneration in MCI. In the hallmark publication by Braak and Braak (1991), the pre- and post-central sulci/gyri are relatively preserved from both amyloid deposition and neurofibrillary tangle deposition until the most advanced pathological stages of AD [50]. In a few studies, neurodegenerative changes in the pre- and post-CS were evident in patients with AD but not in individuals with aMCI. For example, Bauer and colleagues [51] reported hypometabolism in the pre- and post-CS in patients with AD compared to healthy controls and aMCI. Moreover, Yakushev et al. [52] suggested that pathological changes in these areas appear only in advanced AD. Also, previous reports showed no changes in thickness of the pre- and post-CS [53], or in Pittsburgh compound B uptake in primary sensorimotor areas [16] in patients with aMCI or AD and healthy controls. Overall, studies using different methodologies show converging evidence that the pre- and post-CS are not involved in the neurodegenerative process in individuals with aMCI. These findings are in line with our results showing reliable responses for long time scales (e.g., forward story) in the pre- and post-CS in participants with aMCI but not in healthy controls, suggesting that these regions may be involved in compensation, possibly as a result of dysfunction of other brain regions (such as the temporo-parietal areas).

Previous imaging studies have suggested that degeneration of temporo-parietal areas appears somewhat late in the progression of AD, and is associated with AD in its more moderate stages [13 , 54], although functional changes as observed by FDG-PET may be observed in these areas at much earlier stages, including MCI [55, 56] In our study, aMCI participants already showed functional impairment of the temporo-parietal association areas, suggesting that even at more mild stages of cognitive decline, these regions are impaired in their ability to process complex stimuli [13 , 57]. As described above, our data our consistent with a shift in hierarchy from the temporo-parietal association areas to the pre- and post-CS. Functional decline of the temporo-parietal association areas may even precede structural changes that can be demonstrated by imaging [55, 56]. To address this question, we performed a cortical thickness and brain volumetric analysis in our subject groups. As minor but significant changes were observed between groups in the cortical thickness of some lateral temporal structures, a degree of shifting related to structural change is not excluded (Supplementary Table 1). However, the degree of hierarchical shifting of processing of complex auditory information between participants with aMCI and healthy older controls suggested by our data is not likely explained only by these minor structural differences, and some degree of functional shifting is suggested. This functional shift is also supported by the FCA analysis, which showed functional connectivity between the precuneus and the post-CS in the aMCI group. While our data suggest that functional differences may precede structural changes, the exact relationship between functional and anatomical changes in higher order cognitive areas in individuals with MCI is beyond the scope of this study and should be further explored. Given the heterogeneous nature of MCI as a clinical syndrome [58], the minor structural differences observed in the aMCI group are also reassuring in that the underlying cause of aMCI in this group is likely neurodegeneration and not a confounding clinical syndrome.

The fact that long TRWs are observed in the post-CS in participants with aMCI could possibly reflect an alternative mechanism for information processing. This idea is compatible with the hypothesis introduced by Davis and colleagues [59] who suggested that reduced activity in the occipito-temporal regions is replaced by increased activity in more anterior regions (e.g., pre-and post-central gyri). Davis and colleagues [59] have demonstrated this effect in healthy older adults using a semantic encoding task and a task of visual perception. McCarthy et al. [60] have further shown that the relative shifting of brain activity to more anterior regions is more robust in patients with AD using a visual-motor task. It was suggested that this anterior shift reflects compensation for sensory processing deficits in the occipitotemporal regions in aging. Although the post- and pre- CS have sensory and motor primary functions, which are different from the properties of the areas showing decreased reliability in our study, some previous works suggested that plasticity between brain regions that functionally differ is possible. One of the prominent examples is provided in the study by Merabet et al. [61] that showed sensory areas, which normally represent visual activity, process auditory stimuli in blind people. In addition, it has been suggested that in older adults and patients with AD, plasticity is not necessarily material-specific— some regions of interest maybe activated, regardless of material type [62]. Specifically, McCarthy et al. [60] have also demonstrated such a shift of activity from posterior regions to the pre- and post-central gyri. While compensatory effects are typically seen as increased activation, they may also be manifest as topographical shifts, as that seen in posterior-anterior shift. In our study, shifting of activity from posterior to more anterior brain regions was demonstrated only for the participants with aMCI and not for the healthy older adults, who demonstrated a similar pattern of reliable responses to young adults. The fact that a posterior-anterior shift was observed in this study only in participants with aMCI may be due to the difference in the nature of presented stimuli (i.e., auditory dynamic stimuli in the current study versus episodic retrieval and visual perceptual tasks in the previous studies). Alternatively, since the posterior-anterior shift has not been evaluated using inter-SC previously, it is unclear if this effect will be similar with this analytic approach.

A compensation mechanism is supported by the fact that all participants in both groups could retell the story at the end of the fMRI session, indicating that the participants with aMCI were able to compensate behaviorally despite functional impairment of higher cognitive areas. Since the retelling was not evaluated using a structured assessment tool, it might be argued that the participants with aMCI did not fully comprehend the story. This possibility seems less likely given the observation that participants with aMCI, despite lower memory scores, were not significantly functionally impaired and were able to perform reasonably on tests of immediate and delayed logical memory (although they performed worse than healthy older adults, see Table 1) which require comprehension and recall of a short story, similar to the full story in our paradigm.

Memory and information processing

Participants with aMCI showed significant but inverse correlation relationships between memory function and synchronization of brain responses under the long time scale conditions in the precuneus and post-CS. Namely, better memory function was correlated with more reliable brain responses in the precuneus, whereas in the post-CS the opposite effect was observed, suggesting again that participants with memory impairment employed the post-CS as an alternative region for higher-order processing. In that context, we speculate that the association of more synchronized brain responses in the post-CS with poorer MI might be due to enhanced recruitment of the post-CS in subjects with greater cognitive decline, who possibly cannot rely on other high cognitive areas. This does not suggest, however, that full compensation is achieved behaviorally. The fact that the post-CS correlates inversely with memory function suggests that although this area may have be recruited to perform information processing in the participants with aMCI, it was not able to compensate fully for the functions of high-order brain areas. The idea that memory function and information processing are linked has been suggested in a very recent opinion published by Hasson et al. [63]. The authors concluded that memory could not be separated from online information processing; moreover, memory processes are also organized in a hierarchical manner, in which the time scale of memory-dependent processing gradually increases from the primary sensory areas to higher-order areas. Namely, in studies of auditory stimulus processing, Hasson and colleagues suggested that the same neurons at each hierarchical level process and retain information over time. In keeping with this theory and considering that participants with aMCI in our study were suffering from memory impairment, our results could imply that in addition to changes in information processing, the observed functional shift could also affect processes of memory retention.

Additionally, although our paradigm did not test memory function directly, our results converge with prior findings during encoding and retrieval tasks in participants with MCI. Several studies have reported decreased activation in the medial temporal lobe and frontal areas, specifically in the DLPFC [25, 26], and enhanced activity in the pre-CS [25].

Alteration in functional connectivity

In the aMCI group, the precuneus was found to be functionally connected with the pre- and post-CS. Previous research has indicated that the precuneus has anatomical connections to association areas such as the lateral parietal cortex [64]. Functionally, the precuneus is involved in diverse cognitive functions including episodic memory [64 –66]. Furthermore, the precuneus has shown to be affected by the neurodegenerative process early in the course of AD [67 –69]. Our results suggest that although possibly affected by neurodegenerative processes, some functions of the precuneus are still preserved, as evidenced by its reliable responses in the long time scales in both groups, its connectivity with areas in the sensorimotor cortex, and the significant positive correlation between MI and precuneus brain responses.

In summary, we suggest that complex, but not simple, information processing is impaired in aMCI due to dysfunction of higher association areas that are involved in integrating and processing sensory information. Furthermore, our results suggest that there is a functional shift of auditory information processing in aMCI, with longer, more complex and more contextually rich information being processed in other cortical regions from healthy older adults. Future studies should evaluate these functional alterations in more advanced stages of cognitive decline and its association to anatomical degeneration.

Footnotes

ACKNOWLEDGMENTS

We are grateful to all patients and healthy participants for their involvement. We thank Maurice Sarfati for the effort, time, and good will in telling and recording the story. We thank Tal Gonen, Mikhail Katkov, Moran Ovadia, and Efrat Kliper for assistance with analysis. We thank Yulia Golland for helpful comments on the manuscript. We appreciate the I-CORE Program of the Planning and Budgeting Committee and The Israel Science Foundation (grant no. 51/11).

Authors’ disclosures available online ().

Appendix

References

Petersen

, Doody

, Kurz

, Mohs

, Morris

, Rabins

, Ritchie

, Rossor

, Thal

, Winblad

(2001) Current concepts in mild cognitive impairment. Arch Neurol 58, 1985–1992.

Winblad

, Palmer

, Kivipelto

, Jelic

, Fratiglioni

, Wahlund

, Nordberg

, Backman

, Albert

, Almkvist

, Arai

, Basun

, Blennow

, de

, DeCarli

, Erkinjuntti

, Giacobini

, Graff

, Hardy

, Jack

, Jorm

, Ritchie

, van

, Visser

, Petersen

(2004) Mild cognitive impairment–beyond controversies, towards a consensus: Report of the International Working Group on Mild Cognitive Impairment. J Intern Med 256, 240–246.

Chetelat

, Baron

(2003) Early diagnosis of Alzheimer’s disease: Contribution of structural neuroimaging. Neuroimage 18, 525–541.

Fouquet

, Desgranges

, Landeau

, Duchesnay

, Mezenge

, de la Sayette

, Viader

, Baron

, Eustache

, Chetelat

(2009) Longitudinal brain metabolic changes from amnestic mild cognitive impairment to Alzheimer’s disease. Brain 132, 2058–2067.

Gauthier

, Reisberg

, Zaudig

, Petersen

, Ritchie

, Broich

, Belleville

, Brodaty

, Bennett

, Chertkow

, Cummings

, de

, Feldman

, Ganguli

, Hampel

, Scheltens

, Tierney

, Whitehouse

, Winblad

(2006) Mild cognitive impairment. Lancet 367, 1262–1270.

Stam

, van der

, Pijnenburg

, Scheltens

(2003) EEG synchronization in mild cognitive impairment and Alzheimer’s disease. Acta Neurol Scand 108, 90–96.

Albert

, DeKosky

, Dickson

, Dubois

, Feldman

, Fox

, Gamst

, Holtzman

, Jagust

, Petersen

, Snyder

, Carrillo

, Thies

, Phelps

(2011) 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 7, 270–279.

Petersen

, Caracciolo

, Brayne

, Gauthier

, Jelic

, Fratiglioni

(2014) Mild cognitive impairment: A concept in evolution. J Intern Med 275, 214–228.

Nelson

, O’Connor

(2008) Mild cognitive impairment: A neuropsychological perspective. CNS Spectr 13, 56–64.

10.

Taler

, Phillips

(2008) Language performance in Alzheimer’s disease and mild cognitive impairment: A comparative review. J Clin Exp Neuropsychol 30, 501–556.

11.

Mufson

, Binder

, Counts

, DeKosky

, de Toledo-Morrell

, Ginsberg

, Ikonomovic

, Perez

, Scheff

(2012) Mild cognitive impairment: Pathology and mechanisms. Acta Neuropathol 123, 13–30.

12.

Petersen

(2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194.

13.

Chetelat

, Desgranges

, De La Sayette

, Viader

, Eustache

, Baron

(2002) Mapping gray matter loss with voxel-based morphometry in mild cognitive impairment. Neuroreport 13, 1939–1943.

14.

Forsberg

, Engler

, Almkvist

, Blomquist

, Hagman

, Wall

, Ringheim

, Langstrom

, Nordberg

(2008) PET imaging of amyloid deposition in patients with mild cognitive impairment. Neurobiol Aging 29, 1456–1465.

15.

Fouquet

, Desgranges

, La

, Riviere

, Mangin

, Landeau

, Mezenge

, Pelerin

, de la Sayette

, Viader

, Baron

, Eustache

, Chetelat

(2012) Role of hippocampal CA1 atrophy in memory encoding deficits in amnestic Mild Cognitive Impairment. Neuroimage 59, 3309–3315.

16.

Jack

Jr , Lowe

, Senjem

, Weigand

, Kemp

, Shiung

, Knopman

, Boeve

, Klunk

, Mathis

, Petersen

(2008) 11C PiB and structural MRI provide complementary information in imaging of Alzheimer’s disease and amnestic mild cognitive impairment. Brain 131, 665–680.

17.

Jin

, Li

, Hu

, Qin

, Wang

, Xiang

, Yang

, Lu

, Zhong

(2011) Amnestic mild cognitive impairment: Functional MR imaging study of response in posterior cingulate cortex and adjacent precuneus during problem-solving tasks. Radiology 261, 525–533.

18.

Dickerson

, Sperling

(2008) Functional abnormalities of the medial temporal lobe memory system in mild cognitive impairment and Alzheimer’s disease: Insights from functional MRI studies. Neuropsychologia 46, 1624–1635.

19.

Celone

, Calhoun

, Dickerson

, Atri

, Chua

, Miller

, DePeau

, Rentz

, Selkoe

, Blacker

, Albert

, Sperling

(2006) Alterations in memory networks in mild cognitive impairment and Alzheimer’s disease: An independent component analysis. J Neurosci 26, 10222–10231.

20.

Dickerson

, Salat

, Greve

, Chua

, Rand-Giovannetti

, Rentz

, Bertram

, Mullin

, Tanzi

, Blacker

, Albert

, Sperling

(2005) Increased hippocampal activation in mild cognitive impairment compared to normal aging and AD. Neurology 65, 404–411.

21.

Grady

, McIntosh

, Beig

, Keightley

, Burian

, Black

(2003) Evidence from functional neuroimaging of a compensatory prefrontal network in Alzheimer’s disease. J Neurosci 23, 986–993.

22.

Hamalainen

, Pihlajamaki

, Tanila

, Hanninen

, Niskanen

, Tervo

, Karjalainen

, Vanninen

, Soininen

(2007) Increased fMRI responses during encoding in mild cognitive impairment. Neurobiol Aging 28, 1889–1903.

23.

Johnson

, Schmitz

, Moritz

, Meyerand

, Rowley

, Alexander

, Hansen

, Gleason

, Carlsson

, Ries

, Asthana

, Chen

, Reiman

, Alexander

(2006) Activation of brain regions vulnerable to Alzheimer’s disease: The effect of mild cognitive impairment. Neurobiol Aging 27, 1604–1612.

24.

Mandzia

, McAndrews

, Grady

, Graham

, Black

(2009) Neural correlates of incidental memory in mild cognitive impairment: An fMRI study. Neurobiol Aging 30, 717–730.

25.

Petrella

, Krishnan

, Slavin

, Tran

, Murty

, Doraiswamy

(2006) Mild cognitive impairment: Evaluation with 4-T functional MR imaging. Radiology 240, 177–186.

26.

Sperling

, Bates

, Chua

, Cocchiarella

, Rentz

, Rosen

, Schacter

, Albert

(2003) fMRI studies of associative encoding in young and elderly controls and mild Alzheimer’s disease. J Neurol Neurosurg Psychiatry 74, 44–50.

27.

Trivedi

, Stoub

, Murphy

, George

, deToledo-Morrell

, Shah

, Whitfield-Gabrieli

, Gabrieli

, Stebbins

(2011) Entorhinal cortex volume is associated with episodic memory related brain activation in normal aging and amnesic mild cognitive impairment. Brain Imaging Behav 5, 126–136.

28.

, Kemeny

, Park

, Frattali

, Braun

(2005) Language in context: Emergent features of word, sentence, and narrative comprehension. Neuroimage 25, 1002–1015.

29.

Bosch

, Bartres-Faz

, Rami

, Arenaza-Urquijo

, Fernandez-Espejo

, Junque

, Sole-Padulles

, Sanchez-Valle

, Bargallo

, Falcon

, Molinuevo

(2010) Cognitive reserve modulates task-induced activations and deactivations in healthy elders, amnestic mild cognitive impairment and mild Alzheimer’s disease. Cortex 46, 451–461.

30.

Hasson

, Yang

, Vallines

, Heeger

, Rubin

(2008) A hierarchy of temporal receptive windows in human cortex. J Neurosci 28, 2539–2550.

31.

Lerner

, Honey

, Silbert

, Hasson

(2011) Topographic mapping of a hierarchy of temporal receptive windows using a narrated story. J Neurosci 31, 2906–2915.

32.

Hubel

(1988) Eye, brain, and vision Scientific American Library, New York.

33.

Nasreddine

, Phillips

, Bedirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

(2005) The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53, 695–699.

34.

Oren

, Yogev-Seligmann

, Ash

, Hendler

, Giladi

, Lerner

(2015) The Montreal cognitive assessment in cognitively-intact elderly: A case for age-adjusted cutoffs. J Alzheimers Dis 43, 19–22.

35.

Lezak

(1995) Neuropsychological Assessment. Oxford University Press, Inc., New York.

36.

D’Elia

, Satz

, Uchiyama

, White

(1994) Color Trails Test. Psychological Assessment Resources, Odessa, FL.

37.

Kave

(2005) Standardization and norms for a Hebrew naming test. Brain Lang 92, 204–211.

38.

Kave

(2005) Phonemic fluency, semantic fluency, and difference scores: Normative data for adult Hebrew speakers. J Clin Exp Neuropsychol 27, 690–699.

39.

Stuss

, Levine

(2002) Adult clinical neuropsychology: Lessons from studies of the frontal lobes. Annu Rev Psychol 53, 401–433.

40.

Talairach

JTP

(1988) Co-Planar Stereotaxic Atlas of the Human Brain: 3-D Proportional System: An Approach to Cerebral Imaging. Thieme.

41.

Farbood

, Heeger

, Marcus

, Hasson

, Lerner

(2015) The neural processing of hierarchical structure in music and speech at different timescales. Front Neurosci 9, 157.

42.

Benjamini

, Hochberg

(1995) Controlling the false discovery rate - a practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol 57, 289–300.

43.

Benjamini

, Yekutieli

(2001) The control of the false discovery rate in multiple testing under dependency. Ann Stat 29, 1165–1188.

44.

Genovese

, Lazar

, Nichols

(2002) Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage 15, 870–878.

45.

Dale

, Fischl

, Sereno

(1999) Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage 9, 179–194.

46.

Fischl

, Salat

, Busa

, Albert

, Dieterich

, Haselgrove

, van der Kouwe

, Killiany

, Kennedy

, Klaveness

, Montillo

, Makris

, Rosen

, Dale

(2002) Whole brain segmentation: Automated labeling of neuroanatomical structures in the human brain. Neuron 33, 341–355.

47.

Fischl

, Sereno

, Dale

(1999) Cortical surface-based analysis. II: Inflation, flattening, and a surface-based coordinate system. Neuroimage 9, 195–207.

48.

Gainotti

, Quaranta

, Vita

, Marra

(2014) Neuropsychological predictors of conversion from mild cognitive impairment to Alzheimer’s disease. J Alzheimers Dis 38, 481–495.

49.

Bastin

, Salmon

(2014) Early neuropsychological detection of Alzheimer’s disease. Eur J Clin Nutr 68, 1192–1199.

50.

Braak

, Braak

(1991) Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 82, 239–259.

51.

Bauer

, Carbal

, Greve

, Killiany

(2015) Differentiating between normal aging, mild cognitive impairment, and Alzheimer’s disease with FDG-PET: Effects of normalization region and partial volume correction method. J Alzheimers Dis Parkinsonism 3, 1–9.

52.

Yakushev

, Landvogt

, Buchholz

, Fellgiebel

, Hammers

, Scheurich

, Schmidtmann

, Gerhard

, Schreckenberger

, Bartenstein

(2008) Choice of reference area in studies of Alzheimer’s disease using positron emission tomography with fluorodeoxyglucose-F18. Psychiatry Res 164, 143–153.

53.

Singh

, Chertkow

, Lerch

, Evans

, Dorr

, Kabani

(2006) Spatial patterns of cortical thinning in mild cognitive impairment and Alzheimer’s disease. Brain 129, 2885–2893.

54.

Karas

, Scheltens

, Rombouts

, Visser

, van Schijndel

, Fox

, Barkhof

(2004) Global and local gray matter loss in mild cognitive impairment and Alzheimer’s disease. Neuroimage 23, 708–716.

55.

Chetelat

, Eustache

, Viader

, De La Sayette

, Pelerin

, Mezenge

, Hannequin

, Dupuy

, Baron

, Desgranges

(2005) FDG-PET measurement is more accurate than neuropsychological assessments to predict global cognitive deterioration in patients with mild cognitive impairment. Neurocase 11, 14–25.

56.

Pagani

, De Carli

, Morbelli

, Oberg

, Chincarini

, Frisoni

, Galluzzi

, Perneczky

, Drzezga

, van Berckel

, Ossenkoppele

, Didic

, Guedj

, Brugnolo

, Picco

, Arnaldi

, Ferrara

, Buschiazzo

, Sambuceti

, Nobili

(2015) Volume of interest-based [18F]fluorodeoxyglucose PET discriminates MCI converting to Alzheimer’s disease from healthy controls. A European Alzheimer’s Disease Consortium (EADC) study. Neuroimage Clin 7, 34–42.

57.

Whitwell

, Przybelski

, Weigand

, Knopman

, Boeve

, Petersen

, Jack

Jr (2007) 3D maps from multiple MRI illustrate changing atrophy patterns as subjects progress from mild cognitive impairment to Alzheimer’s disease. Brain 130, 1777–1786.

58.

Ash

, Korczyn

(2011) Is amnestic mild cognitive impairment a useful concept? Aging Health 7, 107–122.

59.

Davis

, Dennis

, Daselaar

, Fleck

, Cabeza

(2008) Que PASA? The posterior-anterior shift in aging. Cereb Cortex 18, 1201–1209.

60.

McCarthy

, Benuskova

, Franz

(2014) The age-related posterior-anterior shift as revealed by voxelwise analysis of functional brain networks. Front Aging Neurosci 6, 301.

61.

Merabet

, Rizzo

, Amedi

, Somers

, Pascual-Leone

(2005) What blindness can tell us about seeing again: Merging neurolasticity and neuroprostheses. Nat Rev Neurosci 6, 71–77.

62.

Reuter-Lorenz

, Lustig

(2005) Brain aging: Reorganizing discoveries about the aging mind. Curr Opin Neurobiol 15, 245–251.

63.

Hasson

U CJ

, Honey

(2015) Hierarchical process memory: Memory as an integral component of information processing. Trends Cogn Sci 19, 304–313.

64.

Cavanna

, Trimble

(2006) The precuneus: A review of its functional anatomy and behavioural correlates. Brain 129, 564–583.

65.

Shapira-Lichter

, Vakil

, Glikmann-Johnston

, Siman-Tov

, Caspi

, Paran

, Hendler

(2012) Inside out: A neuro-behavioral signature of free recall dynamics. Neuropsychologia 50, 2245–2256.

66.

Wagner

, Shannon

, Kahn

, Buckner

(2005) Parietal lobe contributions to episodic memory retrieval. Trends Cogn Sci 9, 445–453.

67.

Bradley

, O’Sullivan

, Soper

, Nagy

, King

, Smith

, Shepstone

(2002) Cerebral perfusion SPET correlated with Braak pathological stage in Alzheimer’s disease. Brain 125, 1772–1781.

68.

Buckner

, Snyder

, Shannon

, LaRossa

, Sachs

, Fotenos

, Sheline

, Klunk

, Mathis

, Morris

, Mintun

(2005) Molecular, structural, and functional characterization of Alzheimer’s disease: Evidence for a relationship between default activity, amyloid, and memory. J Neurosci 25, 7709–7717.

69.

Matsuda

(2001) Cerebral blood flow and metabolic abnormalities in Alzheimer’s disease. Ann Nucl Med 15, 85–92.