Prefrontal Activation During Effortful Processing Differentiates Memory Abilities in Adults with Memory Complaints

Abstract

Background:

Identifying individuals at increased risks for developing Alzheimer’s disease (AD) is crucial for early intervention. Memory complaints are associated with brain abnormalities characteristic of AD in cognitively normal older people. However, the utility of memory complaints for predicting mild cognitive impairment (MCI) or AD onset remains controversial, likely due to the heterogeneous nature of this construct.

Objective:

We investigated whether prefrontal oxygenation changes measured by functional near-infrared spectroscopy (fNIRS) during an arduous cognitive task, previously shown to be associated with the AD syndrome, could differentiate memory abilities among individuals with memory complaints. Episodic memory performance was adopted as a proxy for MCI/AD risks since it has been shown to predict AD progression across stages.

Methods:

Thirty-six adults self-reporting memory complaints in the absence of memory impairment completed a verbal list learning test and underwent a digit n-back paradigm with an easy (0-back) and a difficult (2-back) condition. K-means clustering was applied to empirically derive memory complaint subgroups based on fNIRS-based prefrontal oxygenation changes during the effortful 2-back task.

Results:

Cluster analysis revealed two subgroups characterized by high (n = 12) and low (n = 24) bilateral prefrontal activation during the 2-back but not a 0-back task. The low activation group was significantly less accurate across the n-back task and recalled significantly fewer words on the verbal memory test compared to the high activation group.

Conclusion:

fNIRS may have the potential to differentiate verbal memory abilities in individuals with self-reported memory complaints.

Keywords

Episodic memory fNIRS memory complaints prefrontal cortex

INTRODUCTION

Identifying individuals at risk of developing Alz{-}heimer’s disease (AD), which affects approximately 7% of individuals older than 60 years worldwide [1], is crucial for reducing healthcare costs and improving the mental status and quality of life for the patient and caregiver [2]. Research on the early identification of AD to date has focused on examining whether cerebrospinal fluid and blood specimens, as well as positron emission tomography (PET) and magnetic resonance imaging (MRI) scans, can differentiate AD risks [3]. Studies using PET and MRI have shown that the degree of amyloid burden and glucose hypometabolism and the extent of cortical atrophy (e.g., medial temporal lobe degeneration) can distinguish between normal cognition (NC), mild cognitive impairment (MCI), and AD [4 –6].

In recent years, there has been growing interest in using functional near-infrared spectroscopy (fNIRS) to study the AD syndrome due to the ease of applicability of this method to the aging and neuropsychiatric populations [7 –9]. Utilizing the principle of neurovascular coupling, fNIRS measures cerebral hemodynamic changes in terms of fluctuations in the concentration of oxyhemoglobin ([oxy-Hb]) and deoxyhemoglobin ([deoxy-Hb]) induced by neuronal activity [10]. A recent systematic review of 36 fNIRS studies has shown that (amnestic) MCI and clinical AD are associated with reduced cerebral oxygenation changes (i.e., hypoactivation) compared to NC across cognitive tasks, including a widely used working memory (WM) task known as the n-back task [8]. Importantly, such hypoactivation is specific to the condition of high cognitive demands [11, 12], with the level and spatial extent of hypoactivation observed to follow the severity of memory/cognitive impairment (clinical AD > MCI > NC) [13 –15]. Thus, the fNIRS literature has demonstrated a link between attenuated cerebral (prefrontal) oxygenation changes during effortful processing and AD symptoms.

There is increasing recognition that AD progresses along a continuum, from preclinical stage to MCI and then clinically apparent dementia [16, 17]. While much research on the early detection of AD has focused on MCI [18], there has been a growth of interest in elucidating the preclinical stage since this stage may offer the optimal window for therapeutic success [17]. The preclinical stage is defined as normal cognition in the presence of brain abnormalities characteristic of AD. Subjective memory complaints (in the absence of objective memory impairment) have been proposed to be associated with the preclinical stage due to the detection of AD-associated abnormalities, including elevated neuritic amyloid plaques and hippocampal atrophy, in the brains of individuals with self-reported memory complaints on the whole [19, 20]. Yet, the utility of memory complaints for predicting MCI development remains controversial, likely due to large variability among people endorsing memory complaints [21 –23].

The aim of the present study was to evaluate whether fNIRS-based measures previously shown to be linked with the AD syndrome could differentiate verbal memory abilities in adults with memory complaints. Memory impairment is a hallmark of AD, and many studies have shown that episodic memory, which is commonly assessed by a verbal list learning test, predicts AD progression in both MCI individuals [16, 24] and people who scored in the normal range on neuropsychological tests [25 –27]. Therefore, low episodic memory performance was adopted as a proxy for MCI/AD risks in the present study. As such, we investigated whether reduced prefrontal oxygenation changes during an arduous WM (2-back) task could identify low memory performers among adults endorsing memory complaints.

Given that memory complaints represent a poorly understood heterogeneous construct, we utilized k-means clustering to empirically derive memory complaint subgroups based on fNIRS data. K-means clustering has the advantage of forming subgroups using information from multiple features simultaneously without relying on any arbitrary group membership boundary [28]. Accordingly, this unsupervised machine learning algorithm has been increasingly applied to fNIRS data, often with sample sizes falling between 12 and 25, to identify patterns in data across populations [29, 30]. In the present study, changes in left and right prefrontal [oxy-Hb] during the 2-back task were chosen as the classifiers because a lack of bilateral prefrontal oxygenation changes during a difficult cognitive task has been associated with the AD syndrome [12, 31] (see [8] for review).

MATERIALS AND METHODS

Participants

Thirty-six (14 M, 22 F) right-handed adults (mean age = 62.7 years) recruited through online advertisement took part in this study. Individuals who experienced memory problem or often found themselves forgetful were explicitly invited to participate. No participants had a history of any neurological or psychiatric disorders or were on psychotropic medications. All participants scored at least three on a short memory questionnaire for the Chinese [32], implying a substantial level of memory complaints. Nevertheless, no participants had objective memory impairment, defined as more than 1.5 SDs below the age- and education-adjusted normative mean on either the total learning or 10 min delayed recall score of the Hong Kong List Learning Test (HKLLT) [33].

All participants endorsed independence in functional activities and declared no known history of dementia. They scored below 1.0 on the Clinical Dementia Rating scale [34], implying no evidence of dementia. All individuals had normal or corrected-to-normal vision during the experiment. This study was conducted following the Helsinki Declaration of the World Medical Association Assembly and approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee. Participants provided written informed consent before the study.

Procedures and materials

Eligible participants were invited to the Research Center for Neuropsychology Well-being of the Chinese University of Hong Kong for the study. They took the HKLLT and completed the Geriatric Depression Scale [35] and the Geriatric Anxiety Scale [36] to measure mood symptoms over the last week. Participants also filled out the Perceived Deficits Questionnaire to assess subjective cognitive dysfunction in the past week [37] and the Modified Fatigue Impact Scale to assess the feelings of fatigue and the impact of these feelings during the past four weeks [38]. In addition, participants completed the Pittsburgh Sleep Quality Index to measure sleep quality and disturbance over the past month [39] and the Functional Activities Questionnaire to measure instrumental activities of daily living [40]. For each inventory, the total score was chosen as the dependent variable, with a higher score indicating more severe symptoms or worse functioning. Moreover, participants underwent the n-back paradigm under fNIRS recording.

The Hong Kong List Learning Test (HKLLT)

The HKLLT is a widely used standardized verbal memory test in Hong Kong [33]. In this test, participants were presented 16 Chinese vocabulary items verbally three times. The words were from four semantic categories (vegetables, furniture, country, and family member). Immediately after each presentation, they were instructed to recall as many items as possible in any order. Participants were also asked to freely recall the words after 10 minutes without the list of words given. The numbers of correctly recalled items in the three learning trials and the 10 min delayed recall trial were the dependent variables.

N-back task

The n-back task paradigm was adopted from previous fNIRS studies [11, 12]. The task consisted of a low (i.e., 0-back) and a high (i.e., 2-back) memory load condition. The two conditions were presented in an alternate order, which was counterbalanced across participants. In addition, the two conditions were interleaved with a 30 s fixation period, where participants were instructed to sit still and relax.

On each trial, a single-digit was displayed at the center of the computer screen for 1 s, followed by an interstimulus interval of 1 s. In the 0-back condition, participants were required to press the left mouse button if the digit was 0 (target) and press the right button for any other digit (nontarget). In the 2-back condition, participants were asked to press the left button if the current digit was the same as the digit presented two trials back (target) and press the right button if it was not (nontarget). The response time limit was 2 s. Each task block began with a 5 s instruction cue, followed by 20 trials (5 targets and 15 nontargets). Therefore, each task block lasted 45 s, and the entire task paradigm lasted 480 s. The accuracy and mean reaction time (RT) in each condition were the dependent variables.

fNIRS recording

Prefrontal hemodynamic activity during the n-back task was recorded using a continuous-wave 16-channel OEG-SpO2 device (Spectratech Inc., Tokyo, Japan). The device utilizes 770 and 840 nm near-infrared light to estimate changes in [oxy-Hb] and [deoxy-Hb] based on the modified Beer–Lambert law [41]. The sensor consisted of six sources and six detectors alternately arranged in a 2×6 matrix (Fig. 1). The center of the bottom optode row was placed on Fpz. The source–detector separation was 3 cm. Data were sampled at 16 channels at a rate of 12.21 Hz. Notably, we previously used this system and identified reduced prefrontal [oxy-Hb] changes during the 2-back task in MCI individuals compared with cognitively normal older adults [12].

Fig. 1

Optode and channel arrangement and the projected cortical locations for the present functional near-infrared spectroscopy system. Channels were classified into the left and right prefrontal regions of interest.

fNIRS data preprocessing

A custom MATLAB script was used to preprocess fNIRS data. First, raw signals were converted to optical density changes, followed by the application of a 0.10 Hz lower-pass filter with a slope of 60 dB/octave to remove the abrupt spike and cardiac artifacts. The filtered signals were then transformed to changes in [oxy-Hb] and [deoxy-Hb] according to the modified Beer–Lambert law. Slow signal drift was corrected by performing a first-order linear fitting based on the mean values of the 10 s before the start of each task block and the last 10 s of the following fixation period [12]. The correlation-based signal improvement (CBSI) method was applied to remove motion artifacts and systemic activity [42]. Because the [deoxy-Hb] signal mirrored the [oxy-Hb] signal after CBSI, we analyzed only the [oxy-Hb] data.

The corrected [oxy-Hb] data were averaged across time points within each block, condition, and channel for each participant. Some studies have shown that cluster-level variables were more reliable than channel-level variables [43, 44]. Therefore, the [oxy-Hb] data for the seven channels in the left hemisphere (channels 10–16) and the seven channels in the right hemisphere (channels 1–7) were combined to represent the left and right prefrontal regions, respectively [12]. Preprocessing was performed using MATLAB^® R2020a (MathWorks, Natick, MA, USA).

Data analysis

We applied k-means clustering, a classic unsupervised machine learning algorithm that maximizes similarity within clusters and dissimilarity between clusters [28], to empirically divide memory complaint subgroups. Mean changes in left and right prefrontal [oxy-Hb] during the 2-back task were chosen as the two classifiers. The optimal solution was chosen by comparing the average Silhouette scores that represented the quality of clusters created; the Silhouette scores were calculated using the squared Euclidean distance [45]. Next, independent-sample t-tests (or ANOVA) and chi-squared tests were performed to compare demographic features and questionnaire scores between groups. A mixed ANOVA with group as between-subjects factor and condition (0-back, 2-back) and hemisphere (left, right) as with-subjects factors was then conducted to fully analyze PFC activation during the n-back task.

We then examined test performances after subgroup classification. A mixed ANOVA with condition and group as factors was conducted on accuracy and mean RT for the n-back task. Mean RT was calculated after excluding incorrect trials and correct trials where RT was > 2.5 SDs different from the respective mean. In addition, a MANOVA including the number of correct items recalled on the three learning trials and the 10 min delayed recall trial of the HKLLT was conducted to compare verbal memory abilities between groups. Holm–Bonferroni correction was applied for post-hoc tests. All statistical tests were performed using SPSS 26.0 software (IBM Corporation, Armonk, NY, USA). The significance level was set at 0.05, unless otherwise specified.

RESULTS

Cluster analysis

First, k-means clustering that simultaneously considered multiple features (i.e., mean changes in left and right 2-back [oxy-Hb]) was conducted to empirically derive subgroups that exhibited distinct activation patterns during effortful processing. Figure 2 presents the results of the cluster analysis. We observed that the two-cluster solution was optimal because it had a higher average Silhouette score (0.66) compared to solutions with three or more (until sixteen) clusters (≤0.62). According to the two-cluster solution, where 24 and 12 individuals were classified into Cluster 1 and Cluster 2, respectively. Cluster 1 was associated with small increases in [oxy-Hb], whereas Cluster 2 was linked with large increases in [oxy-Hb] consistently in both hemispheres of the PFC. As such, these two clusters were subsequently referred to as low and high activation groups. Table 1 presents the demographic and questionnaire variables of the two groups. No variables significantly differed between groups, ps>0.16.

Fig. 2

K-means clustering with mean changes in oxyhemoglobin concentration ([oxy-Hb]) in the left and right prefrontal regions as the classifiers. A) Average Silhouette values for different cluster sizes. The two-cluster solution had the highest average Silhouette value. B) The Silhouette value for each observation according to the two-cluster solution. All observations had a positive Silhouette value. C) Cluster 1 was associated with small increases in prefrontal [oxy-Hb] (n = 24; low activation group), whereas Cluster 2 was characterized by large increases in [oxy-Hb] across bilateral prefrontal regions (n = 12; high activation group). The centroids of the two clusters were denoted by crosses.

Table 1

Demographics, subjective symptoms, and daily functioning of Cluster 1 (low activation group) and Cluster 2 (high activation group)

	Low activation n = 24		High activation n = 12
	M	SD	M	SD	t/ χ²	p
Demographics
Age (y)	63.6	8.9	60.9	3.7	1.30	0.20
Sex (male/female)	9/15		5/7		0.06	0.81
Education (y)	13.8	3.8	14.3	4.2	0.39	0.70
Subjective symptoms
Abbreviated Memory Inventory for Chinese	4.2	0.7	4.0	0.9	0.63	0.54
Chinese Geriatric Anxiety Scale	6.8	4.7	6.3	4.1	0.26	0.79
Chinese Geriatric Depression Scale	4.5	2.7	3.7	2.9	0.90	0.37
Perceived Deficit Scale^a	22.9	17.2	22.6	13.8	0.06	0.95
Modified Fatigue Impact Scale^a	26.7	16.2	26.6	14.9	0.03	0.98
Daily functioning
Functional Activities Questionnaires	1.6	1.7	0.9	1.7	1.15	0.26
Pittsburgh Sleep Quality Index^a	8.2	3.2	6.6	3.0	1.46	0.16

A chi-squared test was used to compare sex distribution. ^aThree missing data for the low activation group.

Changes in prefrontal [oxy-Hb] during the N-Back Task between groups

Figure 3 presents the mean changes in prefrontal [oxy-Hb] during the n-back task in the two activation groups. A mixed ANOVA with group, condition, and hemisphere as factors was conducted to analyze changes in [oxy-Hb]. We observed significant main effects of condition, F(1, 34) = 78.99, p < 0.001, η $p_{2}^{'}$ = 0.70, and group, F(1, 34) = 29.08, p < 0.001, η $p_{2}^{'}$ = 0.46. The Condition×Hemisphere interaction, F(1, 34) = 9.21, p = 0.005, η $p_{2}^{'}$ = 0.21, and more importantly, the Group×Condition×Hemisphere interaction, F(1, 34) = 4.11-05-202257, p = 0.040, η $p_{2}^{'}$ = 0.12, were also significant. Hence, the effects of group and condition were examined for each hemisphere separately. For both hemispheres, follow-up mixed ANOVA revealed significant results for all the main and interaction effects, ps < 0.001. Post-hoc t-tests with Holm–Bonferroni correction revealed that the two groups significantly differed in mean changes in [oxy-Hb] in both hemispheres only in the 2-back condition, ts > 5.74, ps < 0.001, but not in the 0-back condition, ts < 0.54, ps > 0.59.

Fig. 3

Mean changes in oxyhemoglobin concentration ([oxy-Hb]) in the left and right prefrontal regions during the 0- and 2-back tasks in the high (n = 12) and low (n = 24) activation groups. Asterisks indicate the significance levels of t-test results. Error bars represent the 95% confidence intervals. ^***p < 0.001.

Test performances between groups

Figure 4 presents the test performances for the n-back task and HKLLT in the high and low activation groups. We first analyzed the n-back task (Fig. 4a). For accuracy, the main effects of group, F(1, 34) = 8.89, p = 0.005, η $p_{2}^{'}$ = 0.21, and condition F(1, 34) = 21.39, p < 0.001, η $p_{2}^{'}$ = 0.39, were significant, whereas the Group×Condition interaction was not, F(1, 34) = 2.57, p = 0.12, η $p_{2}^{'}$ = 0.070. Participants were significantly less accurate in the 2-back than 0-back condition, and the low activation group was significantly less accurate than the high activation group overall. For mean RT, only the significant main effect of condition was significant, F(1, 34) = 133.01, p < 0.001, η $p_{2}^{'}$ = 0.80, and no effect involving group was significant, ps > 0.81. Across groups, participants were significantly slower in the 2-back than 0-back condition.

Fig. 4

Behavioral task performance in the high (n = 12) and low (n = 24) activation groups. A) The accuracy and mean RT of the n-back task and (B) the number of correct words recalled in the three learning (i.e., immediate recall; IR) trials and the 10-min delayed recall (DR) trial of the Hong Kong List Learning Test. Asterisks indicate the significance levels of t-test results. Error bars represent the 95% confidence intervals. ^*p < 0.05, ^**p < 0.01.

Next, we examined whether the high and low activation groups significantly differed in verbal memory abilities. Figure 4b shows the number of correct items recalled in the first three learning trials and the 10 min delayed recall trial of the HKLLT. The MANOVA yielded a significant result, F(4, 34) = 3.02, p = 0.033; Wilk’s Λ = 0.72, η $p_{2}^{'}$ = 0.28. Post-hoc t-tests with Holm–Bonferroni revealed that the low activation group recalled significantly fewer words than the high activation group in the first, p = 0.010, second, p = 0.007, but not the third learning trial, p = 0.18. Despite no significant difference in performances in the last learning trial, the low activation group recalled significantly fewer words in the 10 min delayed recall trial compared to the high activation group, p = 0.028.

Unique utility of fNIRS

Finally, we performed a control analysis to determine the unique utility of using fNIRS (i.e., the mean change in 2-back [oxy-Hb]) to differentiate high and low memory performers. Specifically, k-means clustering was applied to 2-back accuracy to evaluate whether similar verbal memory results could be obtained by using a cognitive score. Two memory complaint subgroups that differed in the level of 2-back accuracy (high: n = 28; M = 89.6%, SD = 6.2%; low: n = 8; M = 66.0%, SD = 9.3%) were derived. Unlike using fNIRS to derive subgroups, MANOVA revealed no significant difference between these two subgroups in the HKLLT performance, F(4, 31) = 1.09, p = 0.38; Wilk’s Λ = 0.12, η $p_{2}^{'}$ = 0.12.

DISCUSSION

The aim of the present study was to elucidate whether fNIRS-derived features previously observed to be associated with the AD syndrome could differentiate verbal memory abilities, which are linked with risks of MCI/AD, among adults endorsing memory complaints. Using k-means clustering with mean changes in the left and right [oxy-Hb] during the arduous 2-back task as the classifiers, we identified two subgroups that significantly differed in the degree of bilateral prefrontal activation during the 2-back but not the 0-back task. Importantly, the low activation group performed significantly worse than the high activation group across the n-back task and a standardized verbal list learning test. Our findings suggest that fNIRS can differentiate verbal memory abilities among individuals with self-reported memory complaints.

The existing fNIRS literature has demonstrated that reduced cerebral oxygenation changes during effortful processing are associated with the AD syndrome, with the extent of hypoactivation increasing with the severity of memory/cognitive deficits (clinical AD > MCI > NC) [13 –15]. The present study extends the literature by showing that such fNIRS-derived features can distinguish between individuals with higher and lower (yet normal) levels of memory abilities. Because the two separate groups of participants did not significantly differ in changes in [oxy-Hb] during the easier 0-back task, the distinguishing power appears to be specific to the condition of high WM or cognitive demands. In addition, there is no evidence that demographics, mood and other subjective symptoms, and daily functioning confounded the results due to the lack of significant differences across these aspects between the two groups.

Previous studies have reported mixed findings regarding the relationship between memory complaints and future cognitive decline, raising the possibility that a large amount of heterogeneity in terms of neurocognitive features exist among individuals with memory complaints [21 –23]. One notable strength of the present study is the application of an unsupervised machine learning algorithm to derive memory complaint subgroups based on fNIRS-based measures shown to be linked with the AD syndrome [8]. Most importantly, the two subgroups that we identified differed significantly across working and episodic memory test performances. Thus, the low activation group appears to represent individuals who have lower levels of memory test performance within the cognitively normal population.

Thus, the current study provides initial empirical support for using fNIRS to detect small differences in memory test performance among healthy individuals. Compared to traditional paper-and-pencil memory tests that yield task performance data only, fNIRS generates both behavioral and neuroimaging data within a similar amount of time. Specifically, the administration of the three learning trials and the 10 min delayed recall trial of the HKLLT takes a total of approximately 15 to 20 min. A similar time duration is required to set up the current fNIRS system and administer the n-back task, including instruction and practice. An emerging body of evidence suggests that the combination of behavioral and neuroimaging data allows for more accurate identification of MCI participants than either modality alone [46, 47]. Our findings suggest that fNIRS may have the potential to serve as a quick and comprehensive tool for detecting even small differences in memory abilities among people with memory complaints.

Our finding of the attenuated activation across the bilateral PFC during the arduous 2-back task in individuals with lower levels of memory abilities is consistent with the pivotal role of the PFC and prefrontal-hippocampal circuit in the top-down control of working and episodic memory [48, 49]. Several psychological mechanisms may account for this observation. First, the anterior and lateral PFC has been implicated in various aspects of cognitive control, including task management and planning [50, 51]. Therefore, the lack of prefrontal activation in individuals with lower levels of memory test performance may be due to a reduced readiness or efficiency in deploying compensatory strategies to meet the great cognitive demand of the memory tests. Second, this phenomenon may be attributable to a diminished willingness to expend mental effort to perform a challenging task or a diminished capacity for task engagement due to cognitive overloading among people with a lower level of memory ability.

Our study has some limitations. First, it lacked a longitudinal investigation, which is needed to establish whether reduced changes in prefrontal [oxy-Hb] during a difficult cognitive task predict greater-than-age-expected memory/cognitive decline in individuals with self-reported memory complaints. Second, although we were able to detect significant differences in memory test performances between the two empirically derived subgroups, the present sample size was relatively small. There may be more distinct subgroups among adults with memory complaints, and future work would benefit from increasing the sample size. Third, the present study focused on individuals self-reporting memory complaints. Our findings may not be generalizable to those endorsing other kinds of cognitive complaints.

In conclusion, we identified two separate memory complaint groups linked with distinct memory profiles by applying cluster analysis to fNIRS [oxy-Hb] data. Our findings help to clarify some of the heterogeneity in terms of cognitive test performance among individuals with memory complaints. Although the prognostic value of the results (e.g., predictability of accelerated memory decline or MCI/AD onset) needs to be evaluated by a longitudinal study, the present study provides preliminary support for the use of fNIRS as a potential, sensitive tool for differentiating memory abilities in individuals self-endorsing memory complaints.

Footnotes

ACKNOWLEDGMENTS

The authors would like to thank Christy Cheung, Quin Chan, Samantha Mui, Sophia Sze, and Weltor Hui for their assistance in data collection. Appreciation is extended to all the study participants. This study was supported by the General Research Fund of the Research Grant Council in Hong Kong awarded to ASC (14606519). This study was not preregistered.

Authors’ disclosures available online ().

DATA AVAILABILITY

The dataset analyzed during the current study are available from the corresponding author on reasonable request.

References

Prince

, Wimo

, Guerchet

, Ali

, Wu

, Prina

(2015)World Alzheimer Report 2015. The Global Impact of Dementia: An analysis of prevalence, incidence, cost and trends. Alzheimer’s Disease International, London.

Barnett

, Lewis

, Blackwell

, Taylor

(2014) Early intervention in Alzheimer’s disease: A health economic study of the effects of diagnostic timing. BMC Neurol 14, 101.

Weiner

, Veitch

, Aisen

, Beckett

, Cairns

, Cedarbaum

, Green

, Harvey

, Jack

, Jagust

, Luthman

, Morris

, Petersen

, Saykin

, Shaw

, Shen

, Schwarz

, Toga

, Trojanowski

, Alzheimer’s Disease Neuroimaging Initiative (2015) 2014 Update of the Alzheimer’s Disease Neuroimaging Initiative: A review of papers published since its inception. Alzheimers Dement 11,e1–e120.

Márquez

, Yassa

(2019) Neuroimaging biomarkers for Alzheimer’s disease. Mol Neurodegener 14, 1–14.

Ruan

, D’Onofrio

, Sancarlo

, Bao

, Greco

, Yu

(2016) Potential neuroimaging biomarkers of pathologic brain changes in mild cognitive impairment and Alzheimer’s disease: A systematic review. BMC Geriatr 16, 104.

Varghese

, Sheelakumari

, James

, Mathuranath

(2013) A review of neuroimaging biomarkers of Alzheimer’s disease. Neurol Asia 18, 239–248.

Ehlis

, Schneider

, Dresler

, Fallgatter

(2014) Application of functional near-infrared spectroscopy in psychiatry. Neuroimage 85, 478–488.

Yeung

, Chan

(2020) Functional near-infrared spectroscopy reveals decreased resting oxygenation levels and task-related oxygenation changes in mild cognitive no impairment and dementia: A systematic review. J Psychiatr Res 124, 58–76.

Yeung

, Chan

(2021) A systematic review of the application of functional near-infrared spectroscopy to the study of cerebral hemodynamics in healthy aging. Neuropsychol Rev 31, 139–166.

10.

Villringer

, Chance

(1997) Non-invasive optical spectroscopy and imaging of human brain function. Trends Neurosci 20, 435–442.

11.

Vermeij

, Kessels

, Heskamp

, Simons

, Dautzenberg

, Claassen

(2017) Prefrontal activation may predict working-memory training gain in normal aging and mild cognitive impairment. Brain Imaging Behav 11, 141–154.

12.

Yeung

, Sze

, Woo

, Kwok

, Shum

, Yu

, Chan

(2016) Reduced frontal activations at high working memory load in mild cognitive impairment: Near-infrared spectroscopy. Dement Geriatr Cogn Disord 42, 278–296.

13.

Arai

, Takano

, Miyakawa

, Ota

, Takahashi

, Asaka

, Kawaguchi

(2006) A quantitative near-infrared spectroscopy study: A decrease in cerebral hemoglobin oxygenation in Alzheimer’s disease and mild cognitive impairment. Brain Cogn 61, 189–194.

14.

, Rui

, Chen

, Li

, Schulz

, Zhang

(2018) Early detection of Alzheimer’s disease using non-invasive near-infrared spectroscopy. Front Aging Neurosci 10, 366.

15.

Yap

, Ung

, Ebenezer

, Nordin

, Chin

, Sugathan

, Chan

, Yip

, Kiguchi

, Tang

(2017) Visualizing hyperactivation in neurodegeneration based on prefrontal oxygenation: A comparative study of mild Alzheimer’s disease, mild cognitive impairment, and healthy controls. Front Aging Neurosci 9, 287.

16.

Bondi

, Jak

, Delano-Wood

, Jacobson

, Delis

, Salmon

(2008) Neuropsychological contributions to the early identification of Alzheimer’s disease. Neuropsychol Rev 18, 73–90.

17.

Dubois

, Hampel

, Feldman

, Scheltens

, Aisen

, Andrieu

, Bakardjian

, Benali

, Bertram

, Blennow

, Broich

, Cavedo

, Crutch

, Dartigues

, Duyckaerts

, Epelbaum

, Frisoni

, Gauthier

, Genthon

, Gouw

, Habert

, Holtzman

, Kivipelto

, Lista

, Molinuevo

, O’Bryant

, Rabinovici

, Rowe

, Salloway

, Schneider

, Sperling

, Teichmann

, Carrillo

, Cummings

, Jack Jr CR (2016) Preclinical Alzheimer’s disease: Definition, natural history, and diagnostic criteria. Alzheimers Dement 12, 292–323.

18.

Petersen

, Morris

(2005) Mild cognitive impairment as a clinical entity and treatment target. Arch Neurol 62, 1160–1163.

19.

Kryscio

, Abner

, Cooper

, Fardo

, Jicha

, Nelson

, Smith

, Van Eldik

, Wan

, Schmitt

(2014) Self-reported memory complaints: Implications from a longitudinal cohort with autopsies. Neurology 83, 1359–1365.

20.

van der Flier

, Van Buchem

, Weverling-Rijns-burger

, Mutsaers

, Bollen

, Admiraal-Behloul

, Westendorp

, Middelkoop

(2004) Memory complaints in patients with normal cognition are associated with smaller hippocampal volumes. J Neurol 251, 671–675.

21.

Mitchell

, Beaumont

, Ferguson

, Yadegarfar

, Stubbs

(2014) Risk of dementia and mild cognitive impairment in older people with subjective memory complaints: Meta-analysis. Acta Psychiatr Scand 130, 439–451.

22.

Pearman

, Hertzogs

, Gerstorf

(2014) Little evidence for links between memory complaints and memory performance in very old age: Longitudinal analyses from the Berlin Aging Study. Psychol Aging 29, 828–842.

23.

Yates

, Clare

, Woods

, MRC

CFAS

(2017) Subjective memory complaints, mood and MCI: A follow-up study. Aging Ment Health 21, 313–321.

24.

Gainotti

, Quaranta

, Vita

, Marra

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

25.

Blacker

, Lee

, Muzikansky

, Martin

, Tanzi

, McArdle

, Moss

, Albert

(2007) Neuropsychological measures in normal individuals that predict subsequent cognitive decline. Arch Neurol 64, 862–871.

26.

Oulhaj

, Wilcock

, Smith

, de Jager

(2009) Predicting the time of conversion to MCI in the elderly: Role of verbal expression and learning. Neurology 73, 1436–1442.

27.

Seo

, Kim

, Choi

, Lee

, Choo

(2018) Pre-mild cognitive impairment: Can visual memory predict who rapidly convert to mild cognitive impairment? Psychiatry Investig 15, 869–875.

28.

Hartigan

, Wong

(1979) Algorithm AS 136: A k-means clustering algorithm. J R Stat Soc C Appl Stat 28, 100–108.

29.

, Yu

(2018) Variations of the functional brain network efficiency in a young clinical sample within the autism spectrum: A fNIRS investigation. Front Physiol 9, 67.

30.

Lim

, Ung

, Chan

, Lu

, Sutoko

, Funane

, Kihuchi

, Tang

(2020) A unified analytical framework with multiple fNIRS features for mental workload assessment in the prefrontal cortex. IEEE Trans Neural Syst Rehabil Eng 28, 2367–2376.

31.

Niu

, Li

, Chen

, Ma

, Zhang

(2013) Reduced frontal activation during a working memory task in mild cognitive impairment: A non-invasive near-infrared spectroscopy study. CNS Neurosci Ther 19, 125–131.

32.

Lam

, Lui

, Tam

, Chiu

(2005) Subjective memory complaints in Chinese subjects with mild cognitive impairment and early Alzheimer’s disease. Int J Geriatr Psychiatry 20, 876–882.

33.

Chan

(2006) Hong Kong List Learning Test. 2nd Ed. Department of Psychological and Integrative Neuropsychological Rehabilitation Center, Hong Kong.

34.

Morris

(1997) Clinical dementia rating: A reliable and valid diagnostic and staging measure for dementia of the Alzheimer type. Int Psychogeriatr 9(S1), 173–176.

35.

Lee

HCB

, Chiu

, Kowk

, Leung

(1993) Chinese elderly and the GDS short form: A preliminary study. Clin Gerontol 14, 37–42.

36.

Lin

, Lu

, Gottschling

, Segal

, Tang

(2017) Validation of a Chinese version of the geriatric anxiety scale among community-dwelling older adults in mainland China. J Cross Cult Gerontol 32, 57–70.

37.

Shi

, Wang

, Tian

, Han

, Sha

, Xing

, Yu

(2017) Reliability and validity of Chinese version of perceived deficits questionnaire for depression in patients with MDD. Psychiatry Res 252, 319–324.

38.

Learmonth

, Dlugonski

, Pilutti

, Sandroff

, Motl

(2013) The reliability, precision and clinically meaningful change of walking assessments in multiple sclerosis. Mult Scler 19, 1784–1791.

39.

Tsai

, Wang

, Su

, Yang

, Huang

, Fang

(2005) Psychometric evaluation of the Chinese version of the Pittsburgh Sleep Quality Index (CPSQI) in primary insomnia and control subjects. Qual Life Res 14, 1943–1952.

40.

Pfeffer

, Kurosaki

, Harrah Jr CH , Chance

, Filos

(1982) Measurement of functional activities in older adults in the community. J Gerontol 37, 323–329.

41.

Delpy

, Cope

, van der Zee

, Arridge

, Wray

, Wyatt

(1988) Estimation of optical pathlength through tissue from direct time of flight measurement. Phys Med Biol 33, 1433.

42.

Cui

, Bray

, Reiss

(2010) Functional near infrared spectroscopy (NIRS) signal improvement based on negative correlation between oxygenated and deoxygenated hemoglobin dynamics. Neuroimage 49, 3039–3046.

43.

Plichta

, Herrmann

, Baehne

, Ehlis

, Richter

, Pauli

, Fallgatter

(2006) Event-related functional near-infrared spectroscopy (fNIRS): Are the measurements reliable? Neuroimage 31, 116–124.

44.

Schecklmann

, Ehlis

, Plichta

, Fallgatter

(2008) Functional near-infrared spectroscopy: A long-term reliable tool for measuring brain activity during verbal fluency. Neuroimage 43, 147–155.

45.

Rousseeuw

(1987) Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. J Comput Appl Math 20, 53–65.

46.

Bauer

, Cabral

, Killiany

(2018) Multimodal discrimination between normal aging, mild cognitive impairment and Alzheimer’s disease and prediction of cognitive decline. Diagnostics 8, 14.

47.

Dukart

, Sambataro

, Bertolino

(2016) Accurate prediction of conversion to Alzheimer’s disease using imaging, genetic, and neuropsychological biomarkers. J Alzheimers Dis 49, 1143–1159.

48.

Eichenbaum

(2017) Prefrontal–hippocampal interactions in episodic memory. Nat Rev Neurosci 18, 547–558.

49.

Jin

, Maren

(2015) Prefrontal-hippocampal interactions in memory and emotion. Front Syst Neurosci 9, 170.

50.

Kaller

, Rahm

, Spreer

, Weiller

, Unterrainer

(2011) Dissociable contributions of left and right dorsolateral prefrontal cortex in planning. Cereb Cortex 21, 307–317.

51.

Koechlin

, Basso

, Pietrini

, Panzer

, Grafman

(1999) The role of the anterior prefrontal cortex in human cognition. Nature 399, 148–151.