Correlation Between Prefrontal Functional Connectivity and the Degree of Cognitive Impairment in Alzheimer’s Disease: A Functional Near-Infrared Spectroscopy Study

Abstract

Background:

The development of Alzheimer’s disease (AD) can be divided into subjective cognitive decline (SCD), mild cognitive impairment (MCI), and dementia. Early recognition of pre-AD stages may slow the progression of dementia.

Objective:

This study aimed to explore functional connectivity (FC) changes of the brain prefrontal cortex (PFC) in AD continuum using functional near-infrared spectroscopy (fNIRS), and to analyze its correlation with cognitive function.

Methods:

All participants underwent 48-channel fNIRS at resting-state. Based on Brodmann partitioning, the PFC was divided into eight subregions. The NIRSIT Analysis Tool (v3.7.5) was used to analyze mean ΔHbO2 and FC. Spearman correlation analysis was used to examine associations between FC and cognitive function.

Results:

Compared with HC group, the mean ΔHbO2 and FC were different between multiple subregions in the AD continuum. Both mean ΔHbO2 in the left dorsolateral PFC and average FC decreased sequentially from SCD to MCI to AD groups. Additionally, seven pairs of subregions differed in FC among the three groups: the differences between the MCI and SCD groups were in heterotopic connectivity; the differences between the AD and SCD groups were in left intrahemispheric and homotopic connectivity; whereas the MCI and AD groups differed only in homotopic connectivity. Spearman correlation results showed that FCs were positively correlated with cognitive function.

Conclusions:

These results suggest that the left dorsolateral PFC may be the key cortical impairment in AD. Furthermore, there are different resting-state prefrontal network patterns in AD continuum, and the degree of cognitive impairment is positively correlated with reduced FC strength.

Keywords

Alzheimer’s disease cognitive decline functional connectivity functional near-infrared spectroscopy prefrontal cortex resting-state

INTRODUCTION

Alzheimer’s disease (AD), the most common type of dementia characterized by the core symptoms of progressive cognitive dysfunction and behavioral impairment, is an irreversible and currently incurable neurodegenerative disease known as the “silent threat” [1]. Based on the degree of cognitive impairment and latest biomarker criteria, the progression of AD is often divided into three stages: subjective cognitive decline (SCD), mild cognitive impairment (MCI), and dementia due to AD [2]. SCD refers to an individual’s perceived decline in memory and/or other cognitive abilities relative to their previous level of performance, while objective neuropsychological assessments are normal [3]; it is considered a stage prior to MCI (i.e., the preclinical stage of AD) [4], and is currently the most cutting-edge concept in the field of cognition [5]. MCI is regarded as a transitional stage between normal cognition and dementia (i.e., the early clinical stage of AD) [6], which has a cognitive decline but does not affect the ability of daily living. In contrast to the irreversible stage of AD, early detection and prevention are clinically important and may delay or prevent the onset of dementia.

Functional near-infrared spectroscopy (fNIRS) is a noninvasive neural activity imaging technique based on optical principles that can reveal brain functions in subjects by detecting the hemodynamic changes in brain tissue during resting and task states [7, 8]. Brain functional connectivity (FC) is the dynamic synchronization of neural activity signals between brain regions. The correlation of signals between brain regions under the low-frequency fluctuation of resting state (<0.1 Hz) is called resting state functional connectivity (rs-FC) [9]. Resting state is a natural imaging paradigm that reveals brain functional differences between individuals associated with cognitive impairments. Due to its convenient operating procedure, rs-fNIRS may be more applicable to older adults with cognitive impairment than task-associated fNIRS [10, 11]. Yang et al. [12] showed that quantitative assessment of the resting state of the PFC can identify specific changes in brain function associated with cognitive impairment, which can be used to diagnose and evaluate MCI. Li et al. [8] showed that the brain signal complexity of the AD group was significantly reduced in the default, frontoparietal, ventral and dorsal attention networks as compared to healthy controls; Additionally, Niu et al. [13] showed that the brain’s dynamic FC variability strength (Q) was significantly increased in both amnestic MCI and AD groups as compared to healthy controls, indicating that the brain’s rs-FC was destroyed in amnestic MCI and AD patients. Therefore, fNIRS may be a useful tool for assessing the severity of AD patients, and the brain FC can serve as an effective neural marker. However, most of the current studies have only explored differences in brain FC in MCI and/or AD stages, and few studies have systematically examined the continuum of AD progression across all three stages.

Cognitive functions are achieved by the coordination of multiple brain regions. The prefrontal cortex (PFC), which makes up about 1/3 of the brain, is usually considered to be the cognitive cortex in humans [14]. It is associated with the performance of higher-level neural information processing functions, including memory, executive function, analytical judgement, and reasoning, and is one of the key brain areas currently being studied in the field of cognitive behavior. Additionally, the brain region is not covered with excessive hair, which can reduce scattering and attenuation effects during the experiments. Therefore, we aimed to use rs-fNIRS data in the AD continuum (SCD, MCI, and AD) and healthy controls (HC) to gain a comprehensive understanding of the prefrontal hemodynamic and FC changes in the development of AD and to further evaluate the correlation between the prefrontal FC strength and the degree of cognitive impairment, thereby providing neurobiological markers for early clinical identification of AD patients.

METHODS

Participants

A total of 337 subjects with complaints of memory loss were recruited from the Neurology Department in Longhua Hospital, which is affiliated with Shanghai University of Traditional Chinese Medicine. Meanwhile, 80 healthy volunteers were recruited as the HC group. All participants were right-handed with normal vision and hearing and provided written informed consent prior to the study. The study was approved by the Ethics Committee of Longhua Hospital affiliated to Shanghai University of Traditional Chinese Medicine (No: 2021LCSY118), and the study was registered in the Chinese Clinical Trial Registry (Registration No. ChiCTR2300067418). Figure 1 shows the flow chart for screening subjects.

Fig. 1

Flow chart for screening subjects.

The inclusion criteria for SCD patients were based on the SCD plus framework published in 2014 [15] and included: 1) subjective memory decline, accompanied by self-reported cognitive concerns, which was confirmed by informants; and 2) normal performance on neuropsychologicaltests.

The inclusion criteria for MCI patients were based on the diagnostic criteria proposed by Petersen et al. [16] and included: 1) objective memory impairment confirmed by others as having difficulties in memory; 2) Mini-Mental State Examination (MMSE) score of 24-30; 3) Clinical Dementia Rating (CDR) score of 0.5; 4) Activity of Daily Living (ADL) score < 16; 5) Hachinski Ischemic Scale (HIS) score≤4, Hamilton Depression Scale (HAMD) score≤12; and 6) the absence of dementia.

The inclusion criteria for AD patients were based on the published criteria from National Institute on Aging-Alzheimer’s Association [17] and included: 1) a clear history of cognitive deterioration, meeting the criteria for dementia; 2) MMSE adjusted for education: < = 22 for illiteracy,< = 23 for secondary education,< = 26 for university education; 3) two or more domains of cognitive impairment; 4) progressive deterioration of memory and other cognitive functions; 5) no impairment of consciousness; 6) impairment of activities of daily living, ADL score≥16; and 7) no significant cerebrovascular disease, HIS score < 7.

The inclusion criteria for HC group were as follows: 1) No self-reported memory decline or cognitive concerns; and 2) normal performance on neuropsychological tests.

If any of the subjects had any of the following, they were excluded from the study: 1) a clear history of stroke; 2) severe depression (HAMD score≥24); 3) traumatic brain trauma; 4) other nervous system diseases that can cause cognitive impairment, such as brain tumors, Parkinson’s disease, encephalitis, and epilepsy; 5) other systemic diseases that can cause cognitive impairment, such as thyroid dysfunction, severe anemia, syphilis, and AIDS; 6) history of psychiatric disorders or congenital mental retardation; and 7) inability to cooperate with study procedures.

Neuropsychological assessment

All participants underwent a formal neuropsychological assessment by a trained neuropsychological technician. Global cognitive function was assessed using the MMSE [18] and Montreal Cognitive Assessment (MoCA) [19] scales. The degree of cognitive impairment was assessed using the CDR scales [20]. In addition, the ADL [21] score was collected to assess the patients’ ability to take care for themselves.

fNIRS data acquisition and preprocessing

A portable, multi-channel NIRSIT continuous wave system [12, 22] (OBELAB Inc., Rep.of Korea) was used to measure the hemoglobin oxygenation (ΔHbO2) and deoxygenation (ΔHbR) signals in the brain PFC. The system has 24 laser sources and 32 light detectors with a sampling rate of 8.138 Hz and uses continuous waves with wavelengths of 780 nm and 850 nm [23]. To obtain the best spatial resolution, a distance of 3 cm was maintained between the laser and the detector, providing 48 channels for data analysis. Prior to the start of the experiment, participants were asked to relax for 10 min to adjust their breathing and stabilize their mood to minimize the effect on the data. All participants were asked to sit in a comfortable chair and avoid unnecessary movement. The duration of the fNIRS resting-state measurement was approximately 5 min.

The NIRSIT Analyis Tool v3.7.5 (OBELAB Inc.) was used to preprocess the fNIRS signals [24]. To eliminate the effects of cardiovascular artefacts and environmental noise, the detected raw optical signals were filtered with high-pass and low-pass filters at 0.005 Hz and 0.1 Hz, respectively. We selected high quality channels with a signal-to-noise ratio greater than 30 before extracting the hemodynamic data to avoid misinterpretation. We then used a modified Beer-Lambert law to convert the light signal into ΔHbO2 and ΔHbR concentration data, with units in milli-mol [25]. Because the ΔHbO2 signal can sensitively reflect changes in regional cerebral blood oxygenation and has a better signal-to-noise ratio than the HbR signal [26], we used only the ΔHbO2 data in this study.

Resting-state functional connectivity analysis

Based on Brodmann’s subdivision, the 48 channels were divided into eight subregions: left and right dorsolateral PFC (DLPFC L/R), left and right ventral lateral PFC (VLPFC L/R), left and right frontopolar PFC (FPC L/R), and left and right orbitofrontal cortex (OFC L/R), as shown in Fig. 2 [22]. We obtained the averaged hemodynamic response within each subregion, making the recordings for each participant a N-by-8 matrix, where N indicated time-domain sample numbers and 8 indicated the number of subregions. Based on the obtained hemodynamic parameters, the weight threshold was set to 0.95 and the rs-FC strength between each pair of brain regions was calculated by the Pearson’s correlation coefficient (r). Thus, an 8×8 correlation matrix would be generated for each participant. Subsequently, Fisher’s r - z transformation was applied to convert these r values to Z-scores to improve normality.

Fig. 2

Locate the 8 Brodmann neuroanatomical areas within the PFC based on the fNIRS device channel selection. The orange dot indicates the centre of the device that has located the frontopolar zone (FPz) of the subject. “Rt” indicates right hemisphere, “Lt” indicates left hemisphere.

To further analyze the spatial attributes of the altered prefrontal FC, we divided the altered FC into three spatially distinct groups: 1) intrahemispheric connectivity, which refers to connectivity between regions in the same hemisphere; 2) homotopic connectivity, which refers to interhemispheric connectivity between homologous regions; and 3) heterotopic connectivity, which refers to connectivity between non-homologous regions.

Statistical analysis

All statistical analyses were performed using IBM SPSS Statistics 26. Demographic and clinical characteristics were compared among the four groups (HC, SCD, MCI, and AD) using one-way ANOVA or Kruskal-Wallis H test for continuous variables and χ²-test or Fisher exact test for the categorical variables. The hemodynamic and prefrontal FC changes in all areas of the PFC were compared among the four groups using the multiple linear regression models to exclude the effects of gender [27], age [28], and the education levels [29], and comparisons were made between the three stages of AD and the HC group (SCD versus HC, MCI versus HC, AD versus HC), as well as among the AD continuum (SCD versus MCI, SCD versus AD, MCI versus AD), with the results reported as regression coefficients β (standard error, SE), t-values and p-values. The association between prefrontal FC and MMSE and MoCA scores was assessed by Spearman’s correlation analysis. All the p-values < 0.05 were considered statistically significant.

RESULTS

Demographic and clinical characteristics

A total of 395 participants were eventually enrolled, including 78 in the HC group, 123 in the SCD group, 142 in the MCI group, and 52 in the AD group. The results showed that there was no significant difference in gender (p > 0.05), but there were significant differences in age and education among the four groups (p < 0.001). The more severe the cognitive impairment, the older age and less educated the patients were, suggesting that advanced age and low education level are risk factors for cognitive impairment.

Neuropsychological assessment showed that there were significant differences in MMSE and MoCA scores among the four groups (p < 0.001). These cognitive assessment scores showed a trend of progressive deterioration from SCD to MCI and from MCI to AD, indicating that as cognitive impairment worsened, there was a pronounced decline in global cognitive function, memory, language, and executive function, as shown in Table 1.

Table 1

Demographic and clinical characteristics of the participants

Characteristics	HC (n = 78)	SCD (n = 123)	MCI (n = 142)	AD (n = 52)	p
Gender, n (%)					0.809
Male	35 (44.9%)	47 (38.2%)	57 (40.1%)	20 (38.5%)
Female	43 (55.1%)	76 (61.8%)	85 (59.9%)	32 (61.5%)
Age, n (%)					< 0.001
<65 y	57 (73.1%)	85 (69.1%)	62 (43.7%)	7 (13.5%)
≥65 y	21 (26.9%)	38 (30.9%)	80 (56.3%)	45 (86.5%)
Education (y)	13 (10.75,16)	16 (12,16)	10 (8.75,12)	9 (5,12)	< 0.001
MMSE	30 (29,30)	30 (29,30)	28 (27,29)	21 (17,22)	< 0.001
MoCA	28.5 (27,30)	28 (26,29)	23 (21,25)	14 (9.25,16)	< 0.001

Data expressed as percentage (%) or median (P25, P75). HC, healthy control; SCD, subjective cognitive decline; MCI, mild cognitive impairment; AD, Alzheimer’s disease; MMSE, Mini-Mental State; Examination; MoCA, Montreal Cognitive Assessment.

Hemodynamic changes in the PFC

Table 2 shows the mean ΔHbO2 concentration in the eight subregions among the four groups. Compared with the HC group, the mean ΔHbO2 concentrations in the SCD group were increased in the RFPC, LDLPFC, and LFPC (p < 0.05), and the mean ΔHbO2 concentrations in the MCI group were increased in the RFPC and LFPC (p < 0.05). Furthermore, compared with the SCD group, the mean ΔHbO2 concentrations in the MCI and AD groups were significantly lower (p < 0.05), with the mean and standard deviations of the SCD group being –0.51±1.32, the MCI group being –0.15±1.16, and the AD group being –0.01±0.77. However, the difference between the MCI and AD groups was not significant. In addition, there were no significant differences in mean ΔHbO2 concentrations between the remaining brain regions. Figure 3 shows the hemodynamic changes in the resting-state among the four groups.

Table 2

Differences in mean ΔHbO2 concentrations among the four groups (×10^-6mill-mol)

Subregions	HC (n = 78)	SCD (n = 123)	MCI (n = 142)	AD (n = 52)	p
Right side
RDLPFC	–0.10±0.81	–0.45±1.60	–0.32±1.29	–0.32±1.26	0.039
RVLPFC	–0.15±1.21	–0.41±1.58	–0.50±1.45	–0.27±1.53	0.311
RFPC	0.09±0.67	–0.33±1.15^ΔΔ	–0.50±1.17^ΔΔ	–0.19±1.12	0.001
ROFC	–0.02±0.84	–0.08±1.35	–0.11±1.19	0.13±0.81	0.147
Left side
LDLPFC	–0.09±0.61	–0.51±1.32^ΔΔ	–0.15±1.16^*	–0.01±0.77^*	0.004
LVLPFC	–0.40±1.48	–0.40±1.57	–0.21±1.73	–0.01±1.54	0.245
LFPC	–0.02±0.72	–0.51±1.30^ΔΔ	–0.42±1.25^ΔΔ	–0.35±1.17	0.056
LOFC	0.12±0.83	0.16±1.17	0.02±1.23	0.53±1.06	0.905

Data expressed as mean±standard deviation. Model adjusted for gender, age, and education. Compared with the HC group, ^Δp<0.05, ^ΔΔp<0.01; compared with the SCD group, ^*p<0.05, ^**p<0.01; compared with the MCI group, ^#p<0.05, ^##p<0.01.

Fig. 3

Prefrontal activation maps for all four groups. a) Individual prefrontal activation maps for all four groups. b) Group mean prefrontal activation maps for all four groups.

Functional connectivity analysis in the PFC

Figure 4a shows the FC matrices for all four groups. Red and blue colors represent positive and negative FC, respectively. The results showed that the strength of the average FC decreased sequentially in the HC, SCD, MCI, and AD groups, with the mean and standard deviations of the HC group being 0.60±0.22, the SCD group being 0.51±0.21, the MCI group being 0.47±0.24, and the AD group being 0.40±0.21. Compared with the HC group, the strength of the average FC were significantly lower in the SCD, MCI and AD groups (p < 0.01). Furthermore, compared with the SCD group, the strength of the average FC was significantly lower in the AD group (p < 0.05)(Fig. 4b).

Fig. 4

Functional connectivity analysis in the PFC. a) Maps of Average FC Matrices for all Groups. b) Differences in average FC among the four groups, ^*p < 0.05, ^**p < 0.01.

Functional connectivity analysis between the three stages of AD and the HC group

Based on spatial patterns and eight subregions of the PFC, the differences in FC between the three stages of AD and the HC group were explored separately. The results showed that there were significant differences in left and right intrahemispheric, homotopic and heterotopic connectivity in the SCD, MCI, and AD groups compared with the HC group (p < 0.05). Compared with the HC group, the FC strength in the SCD group was reduced in 13 connectivity pairs, including LDLPFC to LFPC, LFPC to LOFC, RDLPFC to RFPC, RDLPFC to ROFC, LDLPFC to RDLPFC, and LDLPFC to ROFC, etc. (p < 0.05); however, the FC strength was increased in LFPC to LVLPFC and LVLPFC to RFPC (p < 0.05), as shown in Table 3. Compared with the HC group, the FC strength in the MCI group was reduced in 13 connectivity pairs, including LDLPFC to LOFC, RDLPFC to RFPC, LDLPFC to RDLPFC, LFPC to RDLPFC, and LOFC to RDLPFC, etc. (p < 0.05), as shown in Table 4. Compared with the HC group, the FC strength in the AD group was reduced in 15 connectivity pairs, including LDLPFC to LOFC, RDLPFC to RFPC, RDLPFC to ROFC, LFPC to RFPC, LDLPFC to RFPC, and LFPC to ROFC, etc. (p < 0.05), as shown in Table 5.

Table 3

Differences in FCs (p < 0.05) between the SCD and HC groups

Comparison	Spatial patterns	Subregion interaction	B (SE)	t-value	p
SCD - HC	Left hemisphere	LDLPFC - LFPC	–0.174 (0.072)	–2.400	0.017
		LFPC - LOFC	–0.293 (0.075)	–3.885	< 0.001
		LFPC - LVLPFC	0.129 (0.063)	2.052	0.041
	Right hemisphere	RDLPFC - RFPC	–0.237 (0.071)	–3.350	0.001
		RDLPFC - ROFC	–0.274 (0.067)	–4.083	< 0.001
		RFPC - ROFC	–0.152 (0.070)	–2.158	0.032
	Homotopic	LDLPFC - RDLPFC	–0.171 (0.069)	–2.466	0.014
		LFPC - RFPC	–0.192 (0.074)	–2.583	0.010
		LDLPFC - RFPC	–0.137 (0.067)	–2.057	0.040
	Heterotopic	LDLPFC - ROFC	–0.170 (0.067)	–2.532	0.012
		LFPC - RDLPFC	–0.193 (0.067)	–2.859	0.004
		LFPC - ROFC	–0.241 (0.070)	–3.442	0.001
		LOFC - RDLPFC	–0.194 (0.061)	–3.177	0.002
		LOFC - RFPC	–0.161 (0.068)	–2.390	0.017
		LVLPFC - RFPC	0.177 (0.059)	2.976	0.003

SE, standard error. Model adjusted for gender, age, and education.

Table 4

Differences in FCs (p < 0.05) between the MCI and HC groups

Comparison	Spatial patterns	Subregion interaction	B(SE)	t-value	p
MCI - HC	Left hemisphere	LDLPFC - LOFC	–0.208 (0.070)	–2.979	0.003
		LFPC - LOFC	–0.200 (0.077)	–2.605	0.010
	Right hemisphere	RDLPFC - RFPC	–0.284 (0.072)	–3.943	< 0.001
		RDLPFC - ROFC	–0.304 (0.068)	–4.446	< 0.001
		RFPC - ROFC	–0.186 (0.072)	–2.595	0.010
		ROFC - RVLFPC	–0.147 (0.058)	–2.515	0.012
	Homotopic	LDLPFC - RDLPFC	–0.222 (0.071)	–3.319	0.002
		LOFC - ROFC	–0.167 (0.080)	–2.082	0.038
	Heterotopic	LFPC - RDLPFC	–0.225 (0.069)	–3.285	0.001
		LFPC - ROFC	–0.182 (0.071)	–2.559	0.011
		LOFC - RDLPFC	–0.276 (0.062)	–4.458	< 0.001
		LOFC - RFPC	–0.266 (0.069)	–3.871	< 0.001
		LOFC - RVLFPC	–0.128 (0.058)	–2.207	0.028

SE, standard error. Model adjusted for gender, age, and education.

Table 5

Differences in FCs (p < 0.05) between the AD and HC groups

Comparison	Spatial patterns	Subregion interaction	B (SE)	t-value	p
AD - HC	Left hemisphere	LDLPFC - LOFC	–0.233 (0.093)	–2.514	0.012
		LFPC - LOFC	–0.234 (0.102)	–2.288	0.023
	Right hemisphere	RDLPFC - RFPC	–0.244 (0.096)	–2.548	0.011
		RDLPFC - ROFC	–0.369 (0.091)	–4.063	< 0.001
		RFPC - ROFC	–0.248 (0.095)	–2.592	0.010
		RFPC - RVLFPC	–0.197 (0.088)	–2.233	0.026
		LDLPFC - RDLPFC	–0.239 (0.094)	–2.538	0.012
	Homotopic	LFPC - RFPC	–0.274 (0.101)	–2.717	0.007
		LOFC - ROFC	–0.248 (0.107)	–2.328	0.020
	Heterotopic	LDLPFC - RFPC	–0.254 (0.091)	–2.800	0.005
		LDLPFC - ROFC	–0.259 (0.091)	–2.845	0.005
		LFPC - RDLPFC	–0.295 (0.091)	–3.226	0.001
		LFPC - ROFC	–0.262 (0.095)	–2.762	0.006
		LOFC - RDLPFC	–0.222 (0.083)	–2.690	0.007
		LOFC - RFPC	–0.253 (0.092)	–2.766	0.006

SE, standard error. Model adjusted for gender, age, and education.

Functional connectivity analysis based on spatial patterns among the AD continuum

Figure 5 shows all significant differences in FCs among the AD continuum. The results found that there were differences in the left intrahemispheric, homotopic and heterotopic connectivity at the three stages of AD (p < 0.05), while there was no significant difference in the right intrahemispheric (Supplementary Table 1). Figure 5a and 5b shows that differential left intrahemispheric FCs among the three groups were mainly located in the LFPC to LVLPFC, and LOFC to LVLPFC (p < 0.05); Fig. 5c-5e shows that differential heterotopic FCs were mainly located in the LOFC to RFPC, LVLPFC to RDLPFC, and LVLPFC to RFPC (p < 0.05); and Fig. 5f and 5 g shows that differential homotopic FCs were mainly located in the LOFC to ROFC, and LVLPFC to RVLFPC (p < 0.05).

Fig. 5

Differences in FCs among the AD continuum based on spatial patterns.

Functional connectivity analysis based on eight subregions among the AD continuum

Table 6 summarizes the results of the multiple linear regression model for the FC analysis among the AD continuum. Only those with a p-value < 0.05 were selected for presentation. The results showed that there were different FCs of prefrontal subregions between the MCI and SCD groups, the AD and SCD groups, and the MCI and AD groups. Compared with the SCD group, the FC strength in the MCI group was reduced in LOFC to LVLPFC, LOFC to RFPC, LVLPFC to RDLPFC, and LVLPFC to RFPC (p < 0.05); and the FC strength in the AD group was reduced in LFPC to LVLPFC, LOFC to LVLPFC, LOFC to ROFC, LVLPFC to RVLFPC, LVLPFC to RDLPFC, and LVLPFC to RFPC (p < 0.05). However, the FC strength of the AD group was only reduced in LVLPFC to RVLFPC compared with the MCI group (p < 0.05).

Table 6

Differences in FCs (p < 0.05) among the AD continuum based on eight subregions

Subregion interaction	Comparison	B (SE)	t-value	p
LFPC - LVLPFC	Group
	SCD - MCI	0.045 (0.063)	0.713	0.476
	SCD - AD	0.181 (0.088)	2.060	0.040
	MCI - AD	0.136 (0.075)	1.811	0.071
LOFC - LVLPFC	Group
	SCD - MCI	0.117 (0.052)	2.230	0.026
	SCD - AD	0.145 (0.073)	1.981	0.048
	MCI - AD	0.028 (0.063)	0.454	0.650
LOFC - ROFC	Group
	SCD - MCI	0.117 (0.075)	1.573	0.117
	SCD - AD	0.214 (0.104)	2.051	0.041
	MCI - AD	0.097 (0.089)	1.083	0.280
LVLPFC - RVLFPC	Group
	SCD - MCI	0.030 (0.054)	0.556	0.578
	SCD - AD	0.197 (0.075)	2.623	0.009
	MCI - AD	0.167 (0.064)	2.599	0.010
LOFC - RFPC	Group
	SCD - MCI	0.134 (0.063)	2.112	0.035
	SCD - AD	0.134 (0.089)	1.157	0.130
	MCI - AD	0.001 (0.076)	0.009	0.993
LVLPFC - RDLPFC	Group
	SCD - MCI	0.130 (0.064)	2.027	0.043
	SCD - AD	0.211 (0.090)	2.357	0.019
	MCI - AD	0.081 (0.077)	1.601	0.290
LVLPFC - RFPC	Group
	SCD - MCI	0.141 (0.058)	2.409	0.017
	SCD - AD	0.230 (0.082)	2.807	0.005
	MCI - AD	0.089 (0.070)	1.268	0.206

SE, standard error. Model adjusted for gender, age, and education.

Associations between FC strength and MoCA and MMSE scores among the AD continuum

Table 7 shows the correlation between prefrontal FC strength and MoCA and MMSE scores at the three stages of AD. The results demonstrated that the FC strength of LFPC to LVLPFC, LOFC to LVLPFC, LOFC to ROFC, LVLPFC to RVLFPC, and LVLPFC to RFPC were significantly positively correlated with the MMSE and MoCA scores (p < 0.05). However, the FC strength of LOFC to RFPC and LVLPFC to RDLPFC was only significantly positively correlated with the MoCA score (p < 0.01) and not with the MMSE score. The correlation scatter plots between FC strength and MMSE and MoCA scores are shown in Fig. 6 using the LFPC to LVLPFC as an example.

Table 7

Correlation between prefrontal FC strength and MoCA and MMSE scores

Sub -region interaction	MMSE		MoCA
	r	p	r	p
LFPC - LVLPFC	0.16	0.004	0.18	0.002
LOFC - LVLPFC	0.13	0.020	0.16	0.003
LOFC - ROFC	0.14	0.015	0.14	0.012
LVLPFC - RVLFPC	0.13	0.019	0.14	0.012
LOFC - RFPC	0.086	0.124	0.16	0.005
LVLPFC - RDLPFC	0.091	0.107	0.15	0.009
LVLPFC - RFPC	0.19	0.001	0.22	< 0.001

Fig. 6

The correlation scatter plots between FC strength (LFPC to LVLPFC) and MoCA and MMSE scores.

DISCUSSION

This study aimed to investigate the characteristics of prefrontal hemodynamic and FC changes in patients across the spectrum of SCD, MCI, and AD using the fNIRS technique and to explore the association between prefrontal FC and cognitive function. The results of hemodynamic changes showed that the mean ΔHbO2 was significantly different between multiple brain regions at the three stages of AD compared with the HC group, and the mean ΔHbO2 concentration in the LDPFC decreased sequentially in the SCD, MCI, and AD groups. The prefrontal FC results showed that the average FC strength decreased sequentially in the HC, SCD, MCI, and AD groups. Further analysis indicated that there were differences in prefrontal FC across spatial patterns among the four groups, as well as varying degrees of decrease across multiple PFC subregions. In addition, a positive association was found between the severity of cognitive impairment and the reduced FC strength, suggesting that fNIRS may be used as an effective tool for early identification of AD.

Neuronal activity in the brain and cerebral blood flow have a close spatial and temporal relationship. The dynamic communication between neuronal activity and blood vessels in the brain is called neurovascular coupling, which is an important mechanism for fNIRS imaging [30]. In the present study, we assessed the differences in mean ΔHbO2 concentration among the four groups by measuring the hemodynamic response in the PFC. The results found that compared with the HC group, the mean ΔHbO2 concentrations in the SCD group were increased in the RFPC, LDLPFC, and LFPC, and the mean ΔHbO2 concentrations in the MCI group were increased in the RFPC and LFPC, indicating that compensatory neuronal activity occurs in individuals at the early stage of cognitive impairment [31, 32]. The “scaffolding theory of aging and cognition” holds that the aging brain can adapt and compensate for overall neural function and structural decline by increasing prefrontal activation [33]. Further analysis revealed that the mean ΔHbO2 concentration of patients at the three stages of AD differed in the LDLPFC and decreased sequentially among the three groups, indicating that the LDLPFC may be the key cortical impairment in AD patients. With the aggravation of cognitive impairment, the inhibition of brain function occurs. Previous studies have shown that neurovascular dysfunction is closely related to the occurrence and development of AD, which can lead to neurodegeneration and cognitive decline in AD patients [34 –36]. The dorsolateral PFC is mainly involved in working memory, attention, and executive functions, which is closely related to dual-tasking ability and motor control. In addition, the FC of LDLPFC in the resting-state of middle-aged and elderly people is affected by cognitive functions such as memory and execution functions [37].

The results of prefrontal FC showed a sequential decrease in average FC strength among the four groups with the worsening of cognitive impairment in AD patients. Compared with the HC group, the average FC strength of the SCD, MCI, and AD groups were significantly lower, indicating that the brain function of patients with cognitive impairment decreased. Further analysis showed that there was significantly lower in the AD group compared with the SCD group, but there was no significant difference between the MCI and the SCD groups. This finding may be attributed to the fact that the MCI stage is a transitional stage in the progression of AD; the progresses of AD undergoes a quantitative-qualitative change from self-perceived cognitive decline in the SCD stage, then transitioning to objective cognitive decline confirmed by informants in the MCI stage, and finally to severe cognitive impairment accompanied by difficulties in self-care and abnormal mental behavior in the AD stage. Consistent with the present study, previous research also suggests that reduced effective FC between brain regions may be a physiological marker of cognitive impairment [38, 39]. cognitive impairment is strongly associated with reduced prefrontal FC strength in the resting state [40]. These findings indicate that the strength of prefrontal FC may reflect the severity of AD progression.

Based on the distinct spatial patterns in the PFC, we found that there were significant differences in left and right intrahemispheric, homotopic and heterotopic connectivity in patients at the three stages of AD compared with the HC group, which was reflected in the decrease of FC strength in different degrees across multiple PFC subregions. However, there was an increase in the FC strength during the SCD stage, which may be an early compensatory mechanism. The result further demonstrates that the brain network function of patients with cognitive impairment is destroyed. In addition, the intrahemispheric, homotopic and heterotopic connectivity were also differed among the AD continuum. Among them, the difference in intrahemispheric connectivity was mainly reflected in the left side, while there was no significant difference in the right side, suggesting that there is an asymmetry in the impaired FC between the left and right prefrontal cortices in patients with cognitive impairment. The left hemisphere may have more complex behavioral patterns. Previous studies have shown that memory is related to the functional specialization of the brain. The left hemisphere (dominant hemisphere) is mainly associated with literal memory and is responsible for verbal and analytical thinking skills, while the right hemisphere (non-dominant hemisphere) is associated with non-literal memory and is responsible for music, images, and holistic thinking, etc. [41]. Sang et al. [42] showed that there is lateralization in the PFC, with the left PFC playing a greater role at lower brain loads and the right PFC playing a more dominant compensatory role at higher brain loads. Niu et al. [43] found that brain activation in the frontal and temporal lobes on the left side was significantly lower in the MCI group than in the healthy controls, while there was no significant difference on the right side. Yoon et al. [44] also found that the non-amnestic MCI and normal groups showed asymmetric activation of the left and right PFC during the Stroop test, with the right hemisphere more overactivated than the left. Additionally, the differences in homotopic connectivity were mainly in LOFC to ROFC and LVLPFC to RVLPFC, and the strength of FC progressively decreased with the degree of cognitive impairment. Some studies have shown that the OFC is closely related to cognitive functions of emotion appraisal and decision making [45, 46] and the VLPFC is associated with language production and integration of emotional information [47]. Finally, the differences in heterotopic connectivity were located in the LOFC to RFPC, LVLPFC to RDLPFC, and LVLPFC to RFPC. These results demonstrate that the PFC asymmetry is significantly associated with cognitive function and the patients with cognitive impairment show interhemispheric connectivity deficits in the resting state.

Furthermore, the present study showed that compared to the SCD group, the MCI group mainly reflected differences in prefrontal heterotopic connectivity, such as LOFC to RFPC, LVLPFC to RDLPFC and LVLPFC to RFPC; the AD group mainly displayed differences in the left intrahemispheric and homotopic connectivity, including LFPC to LVLPFC, LOFC to LVLPFC, LOFC to ROFC, and LVLPFC to RVLFPC. However, the MCI and AD groups differed only in LVLPFC to RVLFPC. These findings suggest that patients in the AD continuum have different resting-state brain network patterns and that function-specific changes are not uniform across patients, suggesting that the degeneration of their brain function is not synchronous. Some studies have shown [48] that rs-FC can predict AD-related cognitive impairment, from cognitively normal populations to MCI and AD patients. Zhang et al. [49] showed that the FC between the bilateral prefrontal, parietal, occipital, and right temporal lobes was reduced in the MCI group compared to the healthy group. In particular, long-range connections involving the prefrontal and occipital lobes have the potential to be effective biomarkers for identifying MCI. Similarly, Bu et al. [50] found that the level of effective connectivity between brain regions was significantly reduced in the MCI group compared to the healthy control group, such as right prefrontal to left prefrontal cortex, left prefrontal cortex to right occipital lobe, left prefrontal cortex to left occipital lobe, right occipital lobe to left prefrontal cortex, and right prefrontal cortex to left occipital lobe. Taken together, these findings suggest that reduced FCs between brain regions may be potential neurobiomarkers for the clinical identification of AD patients at all stages.

Correlation analysis showed the association between prefrontal FC strength and the degree of cognitive impairment. The MMSE and MoCA scales are the most commonly used neuropsychological tests to assess the characteristics of cognitive function in patients with cognitive impairment. Our study found that the prefrontal FCs were positively correlated with the MMSE and MoCA scores in the AD continuum, indicating that the ability to collaborate between brain regions gradually decreases with the severity of cognitive impairment. The result further demonstrates that the FC differences based on resting-state fNIRS can be used to assess the degree of cognitive impairment in AD patients.

Notably, this study has some limitations. Firstly, our research was a cross-sectional study with four groups of subjects from different populations, which led to some heterogeneity. Thus, future longitudinal population-based rs-fNIRS studies could be refined to validate the findings of this study. Secondly, fewer patients were included in the AD group and they were not evenly distributed across the stages of dementia (mild, moderate, and severe), which reduced the differences between the AD and MCI groups. Therefore, the sample size of participants in the AD group can be further increased in future studies. Finally, the entire fNIRS analysis of this study was based on resting-state data without cognitive tasks, with the aim of exploring the differences in resting-state prefrontal FC during the progression of AD. Hence, one or more cognitive tasks could be further performed to make the findings more convincing at a laterstage.

Conclusion

This fNIRS study indicated that the left dorsolateral PFC may be the key cortical impairment in AD patients. FC strength between different subregions in the PFC decreased sequentially as the degree of cognitive impairment increased. Moreover, patients in the SCD, MCI and AD groups have different resting-state prefrontal network patterns. These findings provided new insights into the pathological mechanisms underlying the development of AD. Besides, fNIRS might be an alternative and reliable tool for early identification of AD.

Footnotes

ACKNOWLEDGMENTS

We would like to thank Ms. Hualing Song for the guidance on statistical methodology.

FUNDING

This work was supported by Medical Innovation Research Program of Shanghai Municipality (Grant No. 21Y11920900), Collaborative Innovation Center project of Pharmaceutical Industry transformation in Hospital of Shanghai Municipality (Grant No. 2093), and Scientific and Technological Innovation Projects of Longhua Hospital (Grant No. CX202052).

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

The data supporting the findings of this study are available on reasonable request from the corresponding author.

The supplementary material is available in the electronic version of this article: .

References

Zvěřová

(2019) Clinical aspects of Alzheimer’s disease. Clin Biochem 72, 3–6.

Jack

Jr , Bennett

, Blennow

, Carrillo

, Dunn

, Haeberlein

, Holtzman

, Jagust

, Jessen

, Karlawish

, Liu

, Molinuevo

, Montine

, Phelps

, Rankin

, Rowe

, Scheltens

, Siemers

, Snyder

HM.

, Sperling R, Contributors (2018) NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement 14, 535–562.

Reisberg

, Ferris

, de Leon

, Crook

(1982) The Global Deterioration Scale for assessment of primary degenerative dementia. Am J Psychiatry 139, 1136–1139.

Hao

, Xi

(2020) Research progresses on imaging of subjective cognitive decline in Alzheimer’s disease. J Tongji Med Univ 41, 811–816.

Han

(2018) Recommendations for diagnosis and treatment of subjective cognitive decline due to preclinical Alzheimer disease in China. J Chin Clin Med Imaging 29, 534–538.

Chinese Society of Dementia and Cognitive Impairment (2022) Chinese expert consensus on the diagnosis and treatment of mild cognitive impairment due to Alzheimer’s disease 2021. Chin J Neurol 55, 421–440.

Uemura

, Shimada

, Doi

, Makizako

, Tsutsumimoto

, Park

, Suzuki

(2016) Reduced prefrontal oxygenation in mild cognitive impairment during memory retrieval. Int J Geriatr Psychiatry 31, 583–591.

Liu

, Zeng

, Huang

, Shang

, Lv

, Wang

, Luo

, Chen

, Shi

, Huang

, Xie

, Chen

(2023) Brain activation during working memory task in amnestic mild cognitive impairment patients and its association with memory and attention. J Alzheimers Dis 91, 863–875.

Zhu

, Xu

, Li

, Peng

, Cai

, Li

, Wu

, Cao

, He

(2017) Decreased functional connectivity and disrupted neural network in the prefrontal cortex of affective disorders: A resting-state fNIRS study. J Affective Disord 221, 132–144.

10.

Wang

, Hu

, Liu

, Li

, Qian

, Niu

(2020) Disrupted functional brain connectivity networks in children with attention-deficit/hyperactivity disorder: Evidence from resting-state functional near-infrared spectroscopy. Neurophotonics 7, 015012.

11.

, Zhu

, Zhao

, Sun

, Wen

, Xie

, Liu

, Niu

, Han

(2018) Decreased resting-state brain signal complexity in patients with mild cognitive impairment and Alzheimer’s disease: A multiscale entropy analysis. Biomed Opt Express 9, 1916–1929.

12.

Yang

, Hong

(2021) Quantitative assessment of resting-state for mild cognitive impairment detection: A functional near-infrared spectroscopy and deep learning approach. J Alzheimers Dis 80, 647–663.

13.

Niu

, Zhu

, Wang

, Li

, Yuan

, Sun

, Han

(2019) Abnormal dynamic functional connectivity and brain states in Alzheimer’s diseases: Functional near-infrared spectroscopy study. Neurophotonics 6, 025010.

14.

Liu

, Chen

(2023) Advances in dynamic metacognitive monitoring theory based on brain science. J Bio-education 11, 1–6.

15.

Jessen

, Amariglio

, van Boxtel

, Breteler

, Ceccaldi

, Chételat

, Dubois

, Dufouil

, Ellis

, van der Flier

, Glodzik

, van Harten

, de Leon

, McHugh

, Mielke

, Molinuevo

, Mosconi

, Osorio

, Perrotin

, Petersen

, Rabin

, Rami

, Reisberg

, Rentz

, Sachdev

, de la Sayette

, Saykin

, Scheltens

, Shulman

, Slavin

, Sperling

, Stewart

, Uspenskaya

, Vellas

, Visser

, Wagner

, Subjective Cognitive Decline Initiative (SCD-I) Working Group (2014) A conceptual framework for research on subjective cognitive decline in preclinical Alzheimer’s disease. Alzheimers Dement 10, 844–852.

16.

Petersen

(2011) Mild cognitive impairment. N Engl J Med 364, 2227–2234.

17.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

18.

, Jia

, Yang

(2016) Mini-Mental State Examination in elderly Chinese: A population-based normative study. J Alzheimers Dis 53, 487–496.

19.

, Li

, Huang

(2012) The Beijing version of the Montreal Cognitive Assessment as a brief screening tool for mild cognitive impairment: A community-based study. BMC Psychiatry 12, 156.

20.

O’Bryant

, Waring

, Cullum

, Hall

, Lacritz

, Massman

, Lupo

, Reisch

, Doody

, Texas Alzheimer’s Research Consortium (2008) Staging dementia using Clinical Dementia Rating Scale Sum of Boxes scores: A Texas Alzheimer’s research consortium study. Arch Neurol 65, 1091–1095.

21.

Perneczky

, Pohl

, Sorg

, Hartmann

, Komossa

, Alexopoulos

, Wagenpfeil

, Kurz

(2006) Complex activities of daily living in mild cognitive impairment: Conceptual and diagnostic issues. Age Ageing 35, 240–245.

22.

Dong

, Choi

, Park

, Baik

, Jung

, Kim

, Lee

(2021) Prefrontal functional connectivity during the verbal fluency task in patients with major depressive disorder: A functional near-infrared spectroscopy study. Front Psychiatry 12, 659814.

23.

Choi

J-K

, Kim

J-M

, Hwang

, Yang

, Choi

M-G

, Bae

H-M

(2016) Time-divided spread-spectrum code-based 400 fW-detectable multichannel fNIRS IC for portable functional brain imaging. IEEE J Solid-State Circuits 2, 484–495.

24.

Tran

, Nguyen

, Boyer

, Fond

, Auquier

, Nguyen

HSA

, Tran

HTN

, Nguyen

, Choi

, Le

, Latkin

, Nathan

, Husain

, McIntyre

, Ho

CSH

, Zhang

MWB

, Ho

RCM

(2023) Differentiating people with schizophrenia from healthy controls in a developing Country: An evaluation of portable functional near infrared spectroscopy (fNIRS) as an adjunct diagnostic tool. Front Psychiatry 14, 1061284.

25.

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, 1433–1442.

26.

, Miao

, Han

, Zeng

, Ouyang

, Yang

, Li

(2017) Complexity analysis of fNIRS signals in ADHD children during working memory task. Sci Rep 7, 829.

27.

Stewart

, Bismark

, Towers

, Coan

, Allen

(2010) Resting frontal EEG asymmetry as an endophenotype for depression risk: Sex-specific patterns of frontal brain asymmetry. J Abnorm Soc Psychol 119, 502–512.

28.

Vermeij

, van Beek

, Olde Rikkert

, Claassen

, Kessels

(2012) Effects of aging on cerebral oxygenation during working-memory performance: A functional near-infrared spectroscopy study. PLoS One 7, e46210.

29.

Baik

, Ko

, Min

, Choi

, Baek

, Kim

, Jung

, Jo

, Lee

, Shin

(2022) Assessment of functional near-infrared spectroscopy by comparing prefrontal cortex activity: A cognitive impairment screening tool. Alzheimer Dis Assoc Disord 36, 266–268.

30.

Fan

, Xie

, Wu

(2022) Advances in functional near infrared spectroscopy in patients with mild cognitive impairment. Chin J Rehabil Med 37, 830–833.

31.

Bookheimer

, Strojwas

, Cohen

, Saunders

, Pericak-Vance

, Mazziotta

, Small

(2000) Patterns of brain activation in people at risk for Alzheimer’s disease. N Engl J Med 343, 450–456.

32.

Yap

, Ung

, Ebenezer

EGM

, 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.

33.

Park

, Reuter-Lorenz

(2009) The adaptive brain: Aging and neurocognitive scaffolding. Annu Rev Psychol 60, 173–196.

34.

Sweeney

, Sagare

, Zlokovic

(2018) Blood–brain barrier breakdown in Alzheimer disease and other neurodegenerative disorders. Nat Rev Neurol 14, 133–150.

35.

Hussain

, Fang

, Chang

(2021) Blood–brain barrier breakdown: An emerging biomarker of cognitive impairment in normal aging and dementia. Front Neurosci 15, 688090.

36.

Zlokovic

(2005) Neurovascular mechanisms of Alzheimer’s neurodegeneration. Trends Neurosci 28, 202–208.

37.

Gao

, Wang

, Qi

, Tang

, Wu

, Zhong

, Zhang

, Ding

, Sun

, Tian

, Nawab

Ali

, Zhu

(2021) Relationship between DLPFC functional connectivity and dual task-walking ability in patients with cognitive impairment. Chin J Rehabil Med 36, 1487–1495.

38.

Yan

, Zhang

, Chen

, Wang

, Liu

(2013) Altered effective connectivity of the default mode network in resting-state amnestic type mild cognitive impairment. J Int Neuropsychol Soc 19, 400–409.

39.

Avirame

, Lesemann

, List

, Witte

, Schreiber

, Flöel

(2015) Cerebral autoregulation and brain networks in occlusive processes of the internal carotid artery. J Cereb Blood Flow Metab 35, 240–247.

40.

Yeung

, Chan

(2020) Functional near-infrared spectroscopy reveals decreased resting oxygenation levels and task-related oxygenation changes in mild cognitive impairment and dementia: A systematic review. J Psychiatr Res 124, 58–76.

41.

Zhu

(2011) Anatomy, function and disorders of the two hemispheres of the brain. Chin J Rehabil Theory Pract 17, 598–600.

42.

Sang

, Wang

, Qiao

(2022) Load-dependent activation of bilateral prefrontal cortex: An f NIRS study. J Army Med Univ 44, 210–216.

43.

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.

44.

Yoon

, Kong

, Choi

, Baek

, Kim

, Shin

, Ko

, Shin

(2019) Neural compensatory response during complex cognitive function tasks in mild cognitive impairment: A near-infrared spectroscopy study. Neural Plast 2019, 7845104.

45.

Kotkowski

, Price

, Franklin

, Salazar

, Woolsey

, DeFronzo

, Blangero

, Glahn

, Fox

(2019) A neural signature of metabolic syndrome. Hum Brain Mapp 40, 3575–3588.

46.

Lopez-Persem

, Roumazeilles

, Folloni

, Marche

, Fouragnan

, Khalighinejad

, Rushworth

MFS

, Sallet

(2020) Differential functional connectivity underlying asymmetric reward-related activity in human and nonhuman primates. Proc Natl Acad Sci U S A 117, 28452–28462.

47.

Ikuhiro

, Yoko

(2016) Right ventrolateral prefrontal cortex involvement in the integration of emotional processing: Parametric mediation analysis of fMRI. Neurosci Lett 615, 92–97.

48.

Lin

, Rosenberg

, Yoo

, Hsu

, O’Connell

, Chun

(2018) Resting-state functional connectivity predicts cognitive impairment related to Alzheimer’s disease. Front Aging Neurosci 10, 94.

49.

Zhang

, Zhu

, Tian

, Jiang

, Li

, Wang

(2022) Early screening model for mild cognitive impairment based on resting-state functional connectivity: A functional near-infrared spectroscopy study. Neurophotonics 9, 045010.

50.

, Huo

, Qin

, Xu

, Wang

, Li

(2019) Effective connectivity in subjects with mild cognitive impairment as assessed using functional near-infrared spectroscopy. Am J Phys Med Rehabil 98, 438–445.