Characteristic activation pattern and network connectivity of brain in type 2 diabetes mellitus patients with mild cognitive impairment: A functional near-infrared spectroscopy study

Abstract

Background

Type 2 diabetes mellitus (T2DM) and mild cognitive impairment (MCI) are prevalent conditions in the aging population, with growing evidence indicating a synergistic detrimental effect on brain function when comorbid. However, the distinct neurofunctional signatures of comorbid T2DM-MCI remain poorly characterized.

Objective

This study aimed to investigate the characteristic brain activation patterns and functional network connectivity in elderly patients with comorbid T2DM-MCI, compared to those with T2DM alone or MCI alone.

Methods

In this cross-sectional study,75 elderly participants (T2DM = 25, MCI = 25, T2DM-MCI = 25) underwent functional near-infrared spectroscopy (fNIRS) during a verbal fluency task, a 2-back task, single walking, and a dual-task (2-back while walking), followed by 8-min resting-state recording. Task-evoked cortical activation and resting-state functional connectivity were analyzed and compared across groups.

Results

During cognitive tasks, the T2DM-MCI group showed significantly reduced activation in prefrontal and motor cortices compared to single-disease groups. Dual-task performance specifically revealed hypoactivation in ventrolateral prefrontal and occipital regions in T2DM-MCI. Resting-state analysis demonstrated globally diminished functional connectivity in T2DM-MCI, particularly within prefrontal-motor networks and interhemispheric connections, whereas no significant differences were found between T2DM and MCI groups alone.

Conclusions

Comorbid T2DM-MCI exhibits a unique dual-pathology profile characterized by concurrent reductions in task-evoked activation and resting-state network connectivity, suggesting compromised neural efficiency from synergistic metabolic and neurodegenerative processes. These impairments may serve as sensitive biomarkers for early detection of diabetes-associated cognitive decline.

Trial registration: The Chinese Clinical Trial Registry (ChiCTR) registration # ChiCTR2400084469 (https://www.chictr.org.cn).

Keywords

Alzheimer's disease diabetes mellitus dual-task functional near-infrared spectroscopy mild cognitive impairment

Introduction

Diabetes has become a global health crisis, posing substantial burdens on healthcare systems worldwide. According to the International Diabetes Federation (IDF), there were over 537 million adults with diabetes worldwide in 2021, of which >90% were type 2 diabetes mellitus (T2DM) cases. This prevalence is projected to reach 783 million by 2045.¹ These patients frequently exhibit dyslipidemia, hypertension, and hyperinsulinemia, which are closely linked to metabolic syndrome that collectively elevate cardiovascular risk.² Meanwhile, the accelerated aging of the population has made cognitive impairment increasingly severe. Mild cognitive impairment (MCI), vascular dementia, and Alzheimer's disease have become major challenges in global public health. As a transition stage between normal aging and dementia, MCI confers substantially higher dementia risk compared to the cognitively healthy individuals.³ Growing attention has been paid to cognitive comorbidities in T2DM patients, particularly MCI. Robust evidence indicates that T2DM is not only a risk factor for MCI but also accelerates dementia conversion.⁴ Diabetic patients face 1.25- to 1.91-fold higher odds of cognitive impairment than their in non-diabetic individuals.⁵ Epidemiological data indicate that the prevalence of MCI among T2DM patients ranges from 19.9%-45.0%.^6,7 Simultaneously, cognitive impairment caused by T2DM markedly impairs quality of life scores domains including autonomy, participation in past/present activities, and social engagement,⁸ creating a vicious cycle that worsens glycemic control and disease outcomes.

Diabetes impairs cognitive function through multiple mechanisms, including glucose toxicity, oxidative stress, insulin resistance, and advanced glycation end products (AGEs) accumulation.⁹ Meanwhile, as a prodromal stage of dementia, MCI is characterized by neuronal loss and synaptic dysfunction, driven primarily by amyloid-β (Aβ) deposition and abnormal tau protein phosphorylation, disrupting the frontoparietal executive network.¹⁰ Notably, the central nervous system (CNS) is also an important target for diabetes. Diabetic patients have a significantly higher risk of developing cognitive impairment and dementia than non-diabetic populations. This compelling link has become a new focus and major challenge and necessitates moving beyond the traditional complication framework to elucidate the neurobiological basis of T2DM patients with MCI (T2DM-MCI). This holds important theoretical and practical value for understanding early pathological mechanisms of diabetes-associated dementia, identifying sensitive biomarkers, and developing effective early intervention strategies.

Accumulating evidence demonstrates reduced prefrontal cortex activation in both T2DM and MCI patients.^11,12 However, current research primarily focuses on T2DM or MCI populations, leaving the functional neuroimaging signatures of comorbid T2DM-MCI poorly characterized. This makes it difficult to clarify the complex interactive mechanisms and specific pathological features of T2DM-MCI. Building on documented independent effects of T2DM and MCI on neural function, we hypothesize that T2DM-MCI exhibits distinct functional brain patterns compared to either condition alone. Here, we integrate dual-task paradigm with functional near-infrared spectroscopy (fNIRS) to investigate differences in task-state and resting-state brain features between T2DM-MCI, T2DM, and MCI groups. It is important to note that the primary objective of this study was not to compare pathological groups against healthy aging, but rather to delineate the specific and interactive neurofunctional signatures of T2DM and MCI by directly contrasting the T2DM-only, MCI-only, and comorbid T2DM-MCI groups. This study offers novel insights into metabolic-cognitive interplay.

Methods

Participants

This study employed a cross-sectional research methodology, conducting participant recruitment in community. A total of 75 elderly participants were enrolled and stratified into: the T2DM group (n = 25), MCI group (n = 25), and T2DM-MCI group (n = 25) based on diagnostic criteria. All participants were provided with written informed consent prior to enrollment. The research protocol was approved by the Ethics Committee of Shandong Sport University (Approval #2023026).We implemented stratified inclusion criteria as follows: (1) T2DM group: Diagnosed with diabetes according to WHO 2019 criteria (FBG ≥7.0 mmol/L); Corrected visual acuity ≥0.8 and normal hearing; Intact cognitive function (Montreal Cognitive Assessment (MoCA) score ≥26); Ability to independently complete experimental tasks; Right-handedness. (2) MCI group: Diagnosed according to International Working Group criteria; MoCA score between 18–26; Normal Activities of Daily Living Scale scores; Absence of severe glucose metabolism abnormalities; Right-handedness. (3) T2DM-MCI group: Concurrent meeting of both T2DM and MCI criteria. Exclusion criteria (applied uniformly): Age >80 years; Severe cognitive impairment (MoCA ≤18); Comorbid malignancies or severe cardiovascular/cerebrovascular diseases; Subjects on medications affecting cognitive function that could not be adjusted; Failed compliance assessment; Concurrent participation in other interventional studies.

During the participant recruitment and fNIRS data acquisition phases, the research staff responsible for operating the NirSmart device and guiding the participants through the experimental tasks were blinded to the group assignments (T2DM, MCI, or T2DM-MCI) of the participants. The group allocation was managed by a separate coordinator who was not involved in the data collection process.

Study design

Procedures

We employed a mixed experimental design with two complementary components, including task-state paradigms and resting-state recordings. The task-state module included a Verbal Fluency Task (VFT) and a dual-task paradigm, both of which have been shown to elicit cortical activation reflecting neural resource allocation.^13,14 VFT task: Each trial consisted of 3 consecutive 20-s blocks. In each 20-s module, participants generated composite words containing the target characters “Da” (big), “Bei” (north), or “Bai” (white) guided by auditory cues, with exemplars including “daxiang” and “beiji”. The rest interval comprised counting aloud numbers 1–5 repeated for 30 s preceding the task phase and 60 s succeeding it. Dual-task paradigm: (1) 2-back task: Computerized 2-back working memory task programmed with E-Prime 3.0 (Psychology Software Tools, USA), presenting 0–9-digit sequences (1000 ms stimulus duration, 1500 ms inter-stimulus interval). Participants held a dual-button response box in their right hand to make real-time judgments, pressing “Y” when the current digit matched the one presented two trials prior, otherwise pressing “N”. Each task block contained 10 pseudo-randomized digits (including 8 valid trials). Practice trials required ≥60% accuracy to proceed to formal testing (non-responses or reaction times >1500 ms counted as errors). (2) Walking task (ST): Subjects were required to perform natural back-and-forth walking on a 5-meter straight-line path at self-paced speed. (3) Dual-task (DT): Simultaneously execute the 2-back cognitive task through Bluetooth headphones during gait movement, with each task phase lasting 30 s. All tasks adopted Block design (30-s rest/30-s task). For resting-state modules, subjects sat quietly for environmental adaptation over 6 min before wearing fNIRS head caps, maintaining fixation on a cross marker 1.5 meters ahead while keeping uniform respiration for 8-min resting-state signal acquisition. Procedure flowchart is illustrated in Figure 1.

Figure 1.

Experimental flowchart. The protocol included an 8-min resting-state recording and a series of task-state paradigms: Verbal Fluency Task (VFT), 2-back task (SC), single walking task (ST), and dual-task (DT).

Data acquisition

Utilized the multichannel dual-wavelength continuous-wave imaging system NirSmart device (Danyang Huichuang Medical Equipment Co., Ltd, Jiangsu, China) to continuously measure and record changes in cerebral oxygenated hemoglobin (HbO₂) and deoxygenated hemoglobin (HbR) concentrations during task performance. The validity of this device has been demonstrated in previous studies.¹⁵The system consisted of near-infrared light sources (light-emitting diodes, LEDs) and avalanche photodiodes (APDs) as detectors. The light source probes employed wavelengths of 730 nm and 850 nm, with a sampling rate of 11 Hz. The experiment utilized 24 light sources and 16 detectors, forming 48 effective channels. The emitter-detector distance averaged 3 cm (range: 2.7–3.3 cm). The primary brain regions observed in this study included the left prefrontal cortex (LPFC), right prefrontal cortex (RPFC), left motor cortex (LMC), right motor cortex (RMC), left occipital cortex (LOC), and right occipital cortex (ROC). The prefrontal cortex was further subdivided into: left frontopolar region (LFPC), right frontopolar region (RFPC), left orbitofrontal cortex (LOFC), right orbitofrontal cortex (ROFC), left dorsolateral prefrontal cortex (LDLPFC), right dorsolateral prefrontal cortex (RDLPFC), right ventrolateral prefrontal cortex (RVLPFC), and left ventrolateral prefrontal cortex (LVLPFC).¹⁶ After fNIRS recording, probe positions were determined using a 3D digitizer, and probabilistic registration methods were applied to align fNIRS channel locations with MNI space coordinates, establishing correspondence with cerebral parcellations (Table 1) (probe arrangement shown in Figure 2).

Figure 2.

fNIRS channel arrangement for different task paradigms. (Left Panel) Channel layout utilized during the Verbal Fluency Task (VFT). (Right Panel) Channel layout utilized during the dual-task paradigm (walking combined with 2-back task). Each numbered circle represents a single measurement channel. The red circles indicate light source optodes (emitters), the blue circles indicate detector optodes, and the gray lines connecting them represent the formed channels. This configuration covers key regions of interest, including the prefrontal cortex (PFC), motor cortex (MC), and occipital cortex (OC), as detailed in Table 1.

Table 1.

fNIRS channel configuration and corresponding regions of interest (ROIs).

ROI	Channels (CH)
Right Ventrolateral Prefrontal Cortex (RVLPFC)	3,4
Left Ventrolateral Prefrontal Cortex (LVLPFC)	11,12
Right Dorsolateral Prefrontal Cortex (RDLPFC)	17,18,19
Left Dorsolateral Prefrontal Cortex (LDLPFC)	23,24,25
Right frontopolar area (RFPA)	7,21
Left frontopolar area (LFPA)	10,20,22
Right Orbitofrontal Region (ROFC)	5,6
Left Orbitofrontal Region (LOFC)	8,9
Right Motor Cortex (RMC)	1,2,15,16,28,29,34,35,42,45
Left Motor Cortex (LMC)	13,14,25,27,30,31,32,33,43,44
Right Occipital Cortex (ROC)	39,40,41,47,48
Left Occipital Cortex (LOC)	36,37,38,46

fNIRS data preprocessing

This study employed the NirSpark functional near-infrared spectroscopy analysis system (HuiChuang Medical, China) for data processing and analysis. All preprocessing steps included: (1) motion artifact correction using a 0.5-s sliding time window to detect signal anomalies, with automatic correction triggered when optical density changes exceeded 6× the global standard deviation or 50% of the intrinsic amplitude; (2) band-pass filtering (0.01–0.2 Hz) to eliminate physiological noise and low-frequency drift; (3) hemoglobin concentration conversion from optical intensity signals based on the modified Beer-Lambert law (units: mmol/L).¹⁷ For task-based data, the dual-task paradigm used a −5 to 0 s baseline for correction, while the VFT task employed pre- and post-task baselines (first 10 s and 5–10 s post-task) to compute feature parameters. Resting-state analysis applied 0.01–0.1 Hz band-pass filtering to calculate Pearson correlation coefficients between HbO₂ and HbR time series across regions of interest (ROIs), followed by Fisher-Z transformation to derive functional connectivity strength metrics. This analytical pipeline has demonstrated robust reliability and validity in prior validation studies.

Data analysis

The fNIRS data preprocessing and initial statistical analysis were performed by researchers who were blinded to the group identities. The data files were coded with anonymous identifiers, and the group information was only reintroduced after the completion of all statistical computations (e.g., after the ANOVA and functional connectivity matrices were generated) for the purpose of result interpretation and group comparison.

Statistical analysis was performed using SPSS 26.0 software, and results were expressed as mean ± standard deviation (mean ± SD). First, the normality of the data from the three groups was assessed using the Shapiro-Wilk test, and the homogeneity of variance was verified using Levene's test. If the homogeneity of variance was met, ANOVA was used to compare the inter-group differences in ΔHbO2 (oxygenated hemoglobin) in the regions of interest (ROIs) during the tasks among the three groups, with post-hoc pairwise comparisons conducted using Tukey's HSD method. If the homogeneity of variance was not met, the non-parametric Kruskal-Wallis H test was used. A one-way ANOVA was performed on the average functional connection strength of the brain networks for each group using NirSpark software, followed by false discovery rate (FDR) correction. The significance level was set at p < 0.05.

Results

Baseline characteristics

The baseline demographic and clinical characteristics of the participants are presented in Table 2. No significant differences were observed in gender distribution, height, or weight between MCI group, T2DM group, and T2DM-MCI group (p > 0.05). All participants completed the baseline assessment with 100% protocol adherence.

Table 2.

Baseline demographic characteristics of participants.

	MCI (n = 25)	T2DM (n = 25)	T2DM-MCI (n = 25)	Statistic	p
Age (years)	67.40 ± 3.84	70.30 ± 6.93	71.82 ± 4.45	F = 1.77	0.19
Sex (male%)	60%	50%	45%	χ²=0.68	0.51
Hight (m)	1.61 ± 0.06	1.65 ± 0.07	1.64 ± 0.07	F = 0.23	0.79
Weight (kg)	60.62 ± 6.78	66.53 ± 6.99	61.11 ± 7.48	F = 2.03	0.15
MoCA	22.30 ± 2.58	27.30 ± 0.48	24.11 ± 3.63	F = 31.65	<0.01

Data are presented as Mean ± Standard Deviation or percentage. Continuous variables (Age, Height, Weight, MoCA) were compared using one-way ANOVA, reporting the F statistic. The categorical variable (Sex) was compared using Chi-square test, reporting the χ² statistic.

Spatial distribution and extent of cortical activation across tasks

Significant differences in cortical activation channels were observed among the T2DM group, MCI group, and T2DM-MCI group during cognitive tasks, walking, and dual-task execution. The T2DM-MCI group showed a significantly reduced cortical activation channels compared to the single-disease groups, particularly in the PFC (prefrontal cortex) and MC (motor cortex). Detailed spatial patterns of these activation differences are presented in Table 3 and Figure 3.

Figure 3.

Spatial patterns of cortical activation during different tasks for each group. Group-averaged maps of oxygenated hemoglobin (ΔHbO₂) concentration changes are displayed for the T2DM group (A), the MCI group (B), and the T2DM-MCI group (C). Each row corresponds to a different task: (1) Verbal Fluency Task (VFT), (2) 2-back task, (3) Single Walking Task (ST), and (4) Dual-Task (DT). The color bar indicates the magnitude of ΔHbO₂ (in mmol/L), with red/yellow colors representing activation (increase in HbO₂) and blue colors representing deactivation (decrease in HbO₂).

Table 3.

Number of activated cortical channels across different tasks and brain regions.

Task	Group	Total number of channels	PFC	MC	OC
VFT	T2DM	3	3	/	/
	MCI	2	2	/	/
	T2DM-MCI	1	1	/	/
2-back	T2DM	37	15	19	3
	MCI	28	11	17	0
	T2DM-MCI	18	3	4	1
ST	T2DM	17	5	12	0
	MCI	3	1	1	1
	T2DM-MCI	6	3	2	1
DT	T2DM	33	15	18	0
	MCI	25	8	15	2
	T2DM-MCI	17	6	9	2

The table summarizes the count of channels showing significant activation (ΔHbO₂) for each group during different tasks. VFT: Verbal Fluency Task; ST: Single Walking Task; DT: Dual-Task; PFC: Prefrontal Cortex; MC: Motor Cortex; OC: Occipital Cortex. “/” indicates that the region was not covered by the channel configuration for that specific task.

Task-evoked cortical activation differences among T2DM, MCI, and T2DM-MCI groups

Compared to patients with either T2DM or MCI alone, those with comorbid T2DM-MCI demonstrated a consistent pattern of reduced cortical activation across multiple cognitive and dual-task paradigms.

During the VFT task (Figure 4A), one-way ANOVA revealed significant main effects of group on ΔHbO₂ in the right dorsolateral prefrontal cortex (RDLPFC, CH13, F(2, 72) = 4.550, p = 0.014), left orbitofrontal cortex (LOFC, CH6, F(2, 72) = 3.296, p = 0.043), and left frontopolar area (LFPA, CH16, F(2, 72) = 4.700, p = 0.012). Post-hoc tests indicated that compared to the T2DM-MCI group, the T2DM group exhibited significantly higher ΔHbO₂ in the RDLPFC and LOFC (both p < 0.05), but significantly lower ΔHbO₂ in the LFPA (p < 0.05). Meanwhile, the MCI group showed significantly higher ΔHbO₂ than the T2DM-MCI group in both the RDLPFC and LFPA (both p < 0.05).

Figure 4.

Inter-group comparisons of cortical activation during cognitive tasks. Bar graphs (mean ± SD) show the contrast in ΔHbO₂ responses for channels that exhibited significant differences among the T2DM, MCI, and T2DM-MCI groups during the (A) Verbal Fluency Task (VFT), (B) 2-back task, and (C) Dual-Task (DT). The channel number (CH) and corresponding brain region (see Table 1) are labeled above each graph. Statistical significance based on one-way ANOVA with post-hoc Tukey's HSD test is indicated: *p < 0.05, **p < 0.01 when compared to the T2DM-MCI group.

In the 2-back task (Figure 4B), significant main effects of group were observed in the following channels: the right motor cortex (RMC) at CH2 (F(2, 72) = 4.288, p = 0.017) and CH28 (F(2, 72) = 3.081, p = 0.047); the right dorsolateral prefrontal cortex (RDLPFC) at CH19 (F(2, 72) = 3.360, p = 0.040); and the left frontopolar area (LFPA) at CH22 (F(2, 72) = 3.699, p = 0.030). Post-hoc analysis demonstrated that the T2DM group had significantly higher ΔHbO₂ than the T2DM-MCI group in the RMC (CH2, CH28) and LFPA (CH22) (all p < 0.05). Similarly, the MCI group showed significantly stronger ΔHbO₂ responses compared to the T2DM-MCI group in the RMC (CH2, CH28), RDLPFC (CH19), and LFPA (CH22) (all p < 0.05). No significant differences were observed between the T2DM and MCI groups alone (all p > 0.05).

Under dual-task conditions (Figure 4C), the ANOVA for the left ventrolateral prefrontal cortex (LVLPFC, CH12) did not reveal a statistically significant main effect of group (F (2, 72) = 2.348, p < 0.01). However, post-hoc tests conducted to explore pairwise differences showed that the T2DM group had significantly higher ΔHbO₂ in the LVLPFC compared to the T2DM-MCI group (p < 0.05). Additionally, the MCI group exhibited significantly higher ΔHbO₂ in the right occipital cortex (ROC, CH36) than the T2DM-MCI group (p < 0.05), although the main effect for CH36 was also not significant (F (2, 72) = 2.255, p = 0.112).

During the single walking task, no significant differences in global activation intensity were found among the three groups. The detailed oxygenated hemoglobin (ΔHbO₂) changes for channels exhibiting significant main effects during the VFT and 2-back tasks, including group means, standard deviations, and corresponding ANOVA statistics, are fully presented in Table 4.

Table 4.

Oxygenated hemoglobin (ΔHbO₂) changes in significant channels during the VFT and dual tasks.

Channels (CH)	T2DM (n = 25)	MCI (n = 25)	T2DM-MCI (n = 25)	F	p
CH2	0.047 ± 0.076^a*	0.055 ± 0.056^b**	0.005 ± 0.061	4.288	0.017
CH6	0.037 ± 0.062^a*	0.003 ± 0.048	0.002 ± 0.052	3.296	0.043
CH12	0.211 ± 0.042^a**	0.098 ± 0.151	0.002 ± 0.092	24.64	<0.01
CH13	0.012 ± 0.029^a*	0.014 ± 0.045^b**	−0.017 ± 0.045	4.550	0.014
CH16	0.010 ± 0.034^a**	0.005 ± 0.043^b*	−0.024 ± 0.049	4.700	0.012
CH19	0.023 ± 0.068	0.034 ± 0.058^b*	−0.008 ± 0.048	3.360	0.040
CH22	0.041 ± 0.061^a**	0.038 ± 0.045^b*	−0.002 ± 0.076	3.699	0.030
CH28	0.046 ± 0.067^a*	0.045 ± 0.059^b*	0.008 ± 0.062	3.081	0.047

Data are presented as Mean ± SD in units of mmol/L. The brain region corresponding to each channel is indicated in parentheses (see Table 1 for abbreviations). Superscript letters indicate significant differences in post-hoc comparisons: “a” significant difference between T2DM-MCI and T2DM groups; “b” significant difference between T2DM-MCI and MCI groups. *p < 0.05, **p < 0.01. p-values are from one-way ANOVA.

Resting-state functional connectivity differences among T2DM, MCI, and T2DM-MCI groups

Whole-brain functional connectivity strength analysis demonstrated significantly reduced HbO₂ in the T2DM-MCI group (0.41 ± 0.17; Figure 5C1) versus the MCI group (0.56 ± 0.20; Figure 5B1) and T2DM group (0.59 ± 0.21; Figure 5A1). Similarly, HbR was markedly attenuated in the T2DM-MCI group (0.13 ± 0.12; Figure 5C2) relative to the MCI group (0.20 ± 0.14; Figure 5B2) and T2DM group (0.19 ± 0.13; Figure 5A2). Additionally, no significant differences in functional connectivity strength were observed between T2DM and MCI groups for either HbO₂ or HbR metrics.

Figure 5.

Global resting-state functional connectivity strength. Bar graphs depict the mean whole-brain functional connectivity strength for the T2DM (A), MCI (B), and T2DM-MCI (C) groups. Connectivity strength was calculated separately based on the temporal correlations of oxygenated hemoglobin (HbO₂, panels A1, B1, C1) and deoxygenated hemoglobin (HbR, panels A2, B2, C2) signals.

Analysis based on HbO₂ showed significantly stronger homologous connectivity in T2DM group versus the T2DM-MCI group within the RPFC, RMC, LOC, and ROC (Figure 6A), along with enhanced functional connectivity in 10 heterologous brain networks (Figure 6B). Meanwhile, the MCI group showed significantly higher connectivity in homologous networks of the RMC and LOC and 5 heterologous networks (Figure 6A, B).

Figure 6.

Strength of homologous and heterologous resting-state functional connections. The bar graphs illustrate the functional connectivity strength. (A, C) show connections within homologous (left-right) brain regions. (B, D) show connections between heterologous (different) brain regions. Analyses were performed based on HbO₂ signals (A, B) and HbR signals (C, D). The brain regions are abbreviated as follows: Prefrontal Cortex (PFC), Motor Cortex (MC), Occipital Cortex (OC); L: Left, R: Right. All presented connections survived a one-way ANOVA with False Discovery Rate (FDR) correction for multiple comparisons. Statistical significance is indicated: *p < 0.05, **p < 0.01.

Analysis based on HbR indicated that the T2DM group exhibited significantly higher connectivity strength than the T2DM-MCI group in homologous networks of the RMC and 4 heterologous networks, while the MCI group demonstrated significantly higher connectivity in homologous networks of the ROC and 4 heterologous networks (Figure 6C, D).

Discussion

Our fNIRS investigation employing dual-task paradigms systematically characterized both task-evoked cortical activation and resting-state functional connectivity profiles across T2DM, MCI, and comorbid T2DM-MCI patients. It was found that comorbid patients exhibited a dual impairment pattern of task-state activation decline and resting-state connectivity reduction. During the task-state, the activation range of their PFC and MC was significantly narrowed, and the ability to allocate neural resources during multi-task processing was decreased; during the resting-state, the whole-brain functional connectivity strength was significantly reduced, particularly in PFC-MC networks. Although patients with T2DM or MCI have abnormalities in specific brain regions, they still retain local compensatory mechanisms,^18,19 contrasting with the widespread deficits in comorbid cases. Our study was designed to elucidate the differential brain activation and connectivity patterns among elderly individuals with T2DM, MCI, and their comorbidity. Although the inclusion of a cognitively and metabolically healthy aged control group would have provided a reference for normal aging, its absence does not preclude the internal validity of our key findings. The pronounced hypoactivation and global functional disconnection observed in the T2DM-MCI group, relative to both single-disease groups, underscores a pathological synergy that cannot be attributed to aging alone. This direct inter-group comparison robustly demonstrates that the comorbid condition represents a distinct neurobiological entity, characterized by a dual-hit on neural efficiency and network integrity, rather than a simple summation of the individual effects seen in T2DM or MCI.

Task-state brain activation patterns in comorbid states

Existing functional neuroimaging studies confirms age-related alterations in cortical activation patterns.²⁰ In addition, age-related hypoactivation frequently accompanied by poor task performance.^21,22 while greater brain activation in older individuals has also been documented.²³ As our groups were age-matched, age-related confounds were minimized in this study. Compared to the VFT task, all groups showed greater cortical recruitment during cognitively demanding 2-back and dual-task conditions, consistent with load-dependent neural engagement. This load-dependent activation pattern suggests compensatory neural mechanisms, aligns with previous research.²⁴ Notably, no significant differences of cortical activation were observed between the T2DM group and the MCI group across all task conditions, which may reflect shared neural pathophysiology in the initial phases of these disorders. More important, the T2DM-MCI group demonstrated both task-related hypoactivation and resting-state functional connectivity disruptions relative to single-disease groups. During the VFT, cognitive tasks, and dual-tasks with high cognitive load, the T2DM-MCI group showed reduced PFC and MC activation compared to the single-disease groups. This impaired multi-tasking capacity suggests that the superimposed effects of metabolic disorders and neurodegenerative pathologies.²⁵ Furthermore, this result also indicates that the comorbid state may lead to a decline in neural resource allocation efficiency, distinct from the mechanisms observed in single-disease conditions.

Task-specific activation patterns also emerged in T2DM-MCI group. During the VFT task, the T2DM group showed significantly higher ΔHbO₂ in the ROFC and LOFC compared to the T2DM-MCI group. This indicates that patients with T2DM alone retained relatively intact neural activation capacity in the orbitofrontal cortex regions. However, when T2DM coexisted with MCI, a significant decline in functional activity occurred in these brain areas. This decline may result in greater impairments in reward processing and emotion regulation in T2DM-MCI patients.²⁶ In addition, the MCI group exhibited higher ΔHbO₂ in the RDLPFC and LFPA compared to the T2DM-MCI group. This indicates that, compared to T2DM-MCI patients, MCI-only patients exhibit superior activation levels in key brain regions associated with working memory, attentional control, and problem-solving.^27,28 Conversely, the T2DM-MCI group showed higher ΔHbO₂ in the RDLPFC than the T2DM group, suggesting underlying functional impairment in this brain region among T2DM patients. Notably, when T2DM coexisted with MCI, abnormal activation emerges in the RDLPFC, potentially representing a neural compensatory mechanism to cope with the dual disease burden. However, the efficacy of this compensatory mechanism and its potential to induce functional disturbances in other neural circuits remains to be elucidated. During the 2-back task performance, the T2DM group exhibited significantly higher ΔHbO₂ in the RMC, LOFC, and LFPA compared to the comorbid group, whereas the MCI group showed significantly higher ΔHbO₂ in the RMC, RDLPFC, and LFPA. These results indicate that patients with a single disease (T2DM or MCI) can maintain working memory through cross-network recruitment between the prefrontal cortex and motor cortex.²⁹ Moreover, comorbidity appears to lead a functional decline in multi-regional collaboration across brain regions. Compared to the 2-back task, during dual-tasking, the T2DM-MCI group exhibited significantly lower ΔHbO₂ in the LVLPFC than the T2DM group, and lower ΔHbO₂ in the ROC than the MCI group. These findings indicate that under higher task demands, all three groups demonstrated insufficient cognitive resource allocation, characterized by reduced whole-brain activation and diminished activation differences between single-disease and comorbid patients. As LVLPFC plays a critical role in suppressing irrelevant stimuli (ignoring environmental noise during walking),³⁰ its insufficient activation may amplify cognitive interference during dual-tasking, manifesting as synchronized deterioration in motor coordination and cognitive coherence. The occipital cortex, as the primary hub for visual-spatial information processing, exhibited higher activation in the MCI group than the T2DM-MCI group. Dual-tasking requires real-time processing of visual-spatial information (e.g., obstacle avoidance). Insufficient ROC activation may increase gait instability and fall risk in T2DM-MCI patients.

Zhang et al. found that T2DM patients exhibited significantly higher PFC activation compared to those with T2DM comorbid with major depression, and the latter group showed reduced activation in the left DLPFC.³¹ Consistent with this pattern, we observed significantly decreased PFC activation coupled with increased DLPFC activation in T2DM-MCI patients when compared to the T2DM-only group. This may indicate that DLPFC is a region where abnormal brain activation occurs in T2DM comorbidities. Takehiko et al. found that both healthy older adults and elderly MCI patients exhibit increased PFC activity during dual-tasking, a result consistent with our current study.³² The brain activation patterns in single-disease groups (T2DM or MCI alone) in this study align with prior research. The distinct brain activation patterns observed in T2DM-MCI patients may stem from the synergistic effects of metabolic dysregulation (e.g., insulin resistance, chronic hyperglycemia)³³ and impaired neurovascular coupling^34,35 under comorbid conditions. This inefficient neural resource allocation in comorbid patients impedes their ability to coordinate concurrent motor and cognitive task execution.³⁶

Characteristics of brain functional connectivity

Previous research has well-documented reduced default mode network (DMN) connectivity in T2DM patients,^37,38 as well as dysregulated prefrontal-parietal network activity in MCI patients.^39,40 However, research on whole-brain connectivity decline patterns in T2DM-MCI comorbidity remains scarce. The results of this study demonstrate that the T2DM-MCI group exhibits lower mean whole-brain functional connectivity strength compared to both the T2DM and MCI groups, while no significant difference in mean connectivity strength is observed between the T2DM and MCI groups alone. This suggests that although T2DM and MCI represent distinct diseases with divergent pathological mechanisms, they may converge in their effects on resting-state brain functional connectivity patterns. Alternatively, at specific disease stages, their effects on functional connectivity might not yet manifest clear differentiation. Critically, comorbidity appears to amplify network dysfunction through synergistic effects (e.g., metabolic dysregulation exacerbating neurodegeneration), leading to a comprehensive decline in global brain network efficiency.

Compared to the T2DM and MCI groups, the T2DM-MCI group exhibited lower functional connectivity strength in homologous brain regions of RMC and LOC. This may reflect a decline in sensory input and motor information processing capacity due to comorbidity.⁴¹ In contrast to the heterologous brain network connectivity results, this study found that the T2DM-MCI group had reduced functional connectivity strength in multiple heterologous brain networks compared to the single-disease groups. This suggests that the comorbidity may alter the connectivity patterns of heterologous brain networks. Further analysis revealed that the differences in heterologous functional connectivity between the comorbidity and the single-disease groups were primarily localized in the parietal-occipital (MC-OC) regions. Aberrant cross-network connectivity between the parietal and occipital lobes may be a core neural mechanism of cognitive decline in T2DM-MCI patients. The MC-OC cross-network connection is a key pathway for sensory information integration and spatial-visual joint encoding, and its functional impairment may involve the decoupling of multimodal sensory integration and higher cognitive functions. Simultaneously, the occipital lobe relies on parietal-mediated attentional modulation to optimize visual processing efficiency and integrate body spatial position information, thereby forming a holistic perception of the environment.⁴²The novel finding of this study is that the weakened connectivity in the comorbid group is not a simple additive effect of individual diseases, but rather exhibits “network-specific impairment,” with preferential damage to the higher-order cognitive-motor integration network (PFC-MC). This may be related to the synergistic effects of T2DM-induced neurovascular unit injury and MCI-related synaptic loss jointly disrupting the frontoparietal circuit.^43,44

In summary, our fNIRS data revealed no significant neurofunctional differences between T2DM-only and MCI-only groups during either tasks or resting-state. This may reflect shared early-stage neuropathological mechanisms (e.g., insulin resistance, vascular impairment, neuroinflammation), active compensatory mechanisms in the brain, and methodological challenges. However, patients with comorbid T2DM-MCI exhibited more pronounced cerebral functional abnormalities, highlighting the severe impact of synergistic pathological effects. These findings provide deeper insights into the complex relationship between T2DM and MCI, emphasize the critical need for early identification of high-risk comorbid populations, and suggest future directions for uncovering more specific early biomarkers through multimodal longitudinal studies.

Conclusions

The comorbid condition of T2DM and MCI may lead to decreased efficiency in neural resource allocation and impaired neurovascular coupling, further exacerbating the decline in executive function and information integration capacity. This comorbid state is characterized by unique patterns of brain activation, and its generalized network disconnection features resemble the early pathology of Alzheimer's disease, suggesting that T2DM combined with MCI likely involves dual mechanisms of metabolic dysregulation and neurodegeneration.

Research prospects

Although this study provides important evidence for understanding the brain functional characteristics of T2DM-MCI patients, several limitations should be noted. First, the relatively small sample size (n = 25 per group) in this single-center cross-sectional design may limit the external validity of the conclusions and preclude investigation of longitudinal changes. Second, the laboratory-based dual-task paradigm exhibits discrepancies compared to real-life scenarios, indicating room for improvement in ecological validity. Future research could focus on developing more ecologically valid assessment methods, such as integrating virtual reality technology to simulate daily living scenarios and exploring the clinical value of interdisciplinary comprehensive management strategies. Notably, the characteristic neural markers identified not only provide an objective basis for early diagnosis but also establish a critical foundation for future exercise interventions based on neural mechanisms. These findings offer significant guidance for improving clinical outcomes in patients with metabolic-cognitive dual disorders.

Footnotes

Acknowledgements

The authors have no acknowledgments to report.

ORCID iDs

Jincheng Han

Qinghui Shang

Ethical considerations

The research protocol was approved by the Ethics Committee of Shandong Sport University (Approval #2023026).

Consent to participate

All participants provided their written informed consent to participate in this study.

Consent for publication

Not applicable.

Author contribution(s)

Zhiyuan Sun: Data curation; Formal analysis; Methodology; Validation; Writing – original draft.

Jincheng Han: Conceptualization; Data curation; Formal analysis; Methodology; Writing – original draft.

Hualei Liu: Investigation; Methodology; Software; Validation.

Haiqing Zeng: Investigation; Methodology; Software; Validation; Visualization.

Ran Li: Formal analysis; Investigation; Software; Validation.

Wei Ji: Investigation; Methodology; Visualization.

Dewei Mao: Conceptualization; Funding acquisition; Investigation; Resources; Validation; Visualization.

Xuewen Tian: Conceptualization; Funding acquisition; Resources; Supervision; Validation; Visualization.

Qinghui Shang: Conceptualization; Data curation; Formal analysis; Validation; Visualization.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0531803) and the 2024 Ministry of Education Humanities and Social Sciences Research Foundation (Youth Foundation Project) (24YJC890031).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

References

Sun

Saeedi

Karuranga

, et al. IDF Diabetes Atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract 2022; 183: 109119.

ElSayed

Aleppo

Aroda

, et al. 2. Classification and diagnosis of diabetes: standards of care in diabetes-2023. Diabetes Care 2023; 46: S19–S40.

Pal

Mukadam

Petersen

, et al. Mild cognitive impairment and progression to dementia in people with diabetes, prediabetes and metabolic syndrome: a systematic review and meta-analysis. Soc Psychiatry Psychiatr Epidemiol 2018; 53: 1149–1160.

Yuan

Wang

. Mild cognitive impairment in type 2 diabetes mellitus and related risk factors: a review. Rev Neurosci 2017; 28: 715–723.

Xue

, et al. Diabetes mellitus and risks of cognitive impairment and dementia: a systematic review and meta-analysis of 144 prospective studies. Ageing Res Rev 2019; 55: 100944.

Zheng

, et al. Prevalence and incidence of mild cognitive impairment in adults with diabetes in the United States. Diabetes Res Clin Pract 2023; 205: 110976.

Chakraborty

Hegde

Praharaj

, et al. Age related prevalence of mild cognitive impairment in type 2 diabetes mellitus patients in the Indian population and association of serum lipids with cognitive dysfunction. Front Endocrinol (Lausanne) 2021; 12: 798652.

Hussenoeder

Conrad

Roehr

, et al. Mild cognitive impairment and quality of life in the oldest old: a closer look. Qual Life Res 2020; 29: 1675–1683.

Hadley

Zhang

Harris-Skillman

, et al. Cognitive decline and diabetes: a systematic review of the neuropathological correlates accounting for cognition at death. J Neurol Neurosurg Psychiatry 2022; 93: 246–253.

10.

Jongsiriyanyong

Limpawattana

. Mild cognitive impairment in clinical practice: a review article. Am J Alzheimers Dis Other Demen 2018; 33: 500–507.

11.

Zhao

Tomita

Dutta

. Functional near-infrared spectroscopy of prefrontal cortex during memory encoding and recall in elderly with type 2 diabetes mellitus. Annu Int Conf IEEE Eng Med Biol Soc 2022; 2022: 3323–3326.

12.

Yang

Huang

Yoo

, et al. Detection of mild cognitive impairment using convolutional neural network: temporal-feature maps of functional near-infrared spectroscopy. Front Aging Neurosci 2020; 12: 141.

13.

Lyu

Zhao

, et al. Gender differences in brain region activation during verbal fluency task as detected by fNIRS in patients with depression. World J Biol Psychiatry 2024; 25: 141–150.

14.

Talamonti

Dupuy

Boudaa

, et al. Prefrontal hyperactivation during dual-task walking related to apathy symptoms in older individuals. PLoS One 2022; 17: e0266553.

15.

Zhao

Wang

, et al. Comparative effectiveness of open and closed skill exercises on cognitive function in young adults: a fNIRS study. Sci Rep 2024; 14: 21007.

16.

Szczepanski

Knight

. Insights into human behavior from lesions to the prefrontal cortex. Neuron 2014; 83: 1002–1018.

17.

Shaaban-Ali

Momeni

Denault

. Clinical and technical limitations of cerebral and somatic near-infrared spectroscopy as an oxygenation monitor. J Cardiothorac Vasc Anesth 2021; 35: 763–779.

18.

Yoon

Kong

Choi

, et al. Neural compensatory response during complex cognitive function tasks in mild cognitive impairment: a near-infrared spectroscopy study. Neural Plast 2019; 2019: 7845104.

19.

van Bussel

Backes

van Veenendaal

, et al. Functional brain networks are altered in type 2 diabetes and prediabetes: signs for compensation of cognitive decrements? The Maastricht study. Diabetes 2016; 65: 2404–2413.

20.

Laguë-Beauvais

Brunet

Gagnon

, et al. A fNIRS investigation of switching and inhibition during the modified Stroop task in younger and older adults. Neuroimage 2013; 64: 485–495.

21.

Causse

Chua

Peysakhovich

, et al. Mental workload and neural efficiency quantified in the prefrontal cortex using fNIRS. Sci Rep 2017; 7: 5222.

22.

Wijeakumar

Magnotta

Spencer

. Modulating perceptual complexity and load reveals degradation of the visual working memory network in ageing. Neuroimage 2017; 157: 464–475.

23.

Yeung

Chan

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

24.

Rypma

Berger

D'Esposito

. The influence of working-memory demand and subject performance on prefrontal cortical activity. J Cogn Neurosci 2002; 14: 721–731.

25.

Liu

, et al. Impaired amyloid β-degrading enzymes in brain of streptozotocin-induced diabetic rats. J Endocrinol Invest 2011; 34: 26–31.

26.

Sharpe

Schoenbaum

. Back to basics: making predictions in the orbitofrontal-amygdala circuit. Neurobiol Learn Mem 2016; 131: 201–206.

27.

, et al. Repeated photobiomodulation induced reduction of bilateral cortical hemodynamic activation during a working memory task in healthy older adults. IEEE J Biomed Health Inform 2023; 27: 2876–2885.

28.

Zhang

Liu

, et al. Prefrontal cortex responses to game rewards and losses in individuals with internet gaming disorder: insights from fNIRS during mobile gameplay. J Behav Addict 2025; 14: 347–360.

29.

Dai

Chang

Huang

, et al. Exercise mode and executive function in older adults: an ERP study of task-switching. Brain Cogn 2013; 83: 153–162.

30.

Hoffman

Jefferies

Lambon Ralph

. Ventrolateral prefrontal cortex plays an executive regulation role in comprehension of abstract words: convergent neuropsychological and repetitive TMS evidence. J Neurosci 2010; 30: 15450–15456.

31.

Zhang

Liu

, et al. Characteristic activation pattern and network connectivity of prefrontal cortex in patients with type 2 diabetes mellitus and major depressive disorder during a verbal fluency task: a functional near-infrared spectroscopy study based on network-based statistic prediction. Neuroendocrinology 2024; 114: 1112–1123.

32.

Doi

Makizako

Shimada

, et al. Brain activation during dual-task walking and executive function among older adults with mild cognitive impairment: a fNIRS study. Aging Clin Exp Res 2013; 25: 539–544.

33.

Manschot

Brands

van der Grond

, et al. Brain magnetic resonance imaging correlates of impaired cognition in patients with type 2 diabetes. Diabetes 2006; 55: 1106–1113.

34.

Bégay

Baumeier

Zimmermann

, et al. The C/EBPβ LIP isoform rescues loss of C/EBPβ function in the mouse. Sci Rep 2018; 8: 8417.

35.

Monteiro

Castro

Pereira

, et al. Neurovascular coupling is impaired in hypertensive and diabetic subjects without symptomatic cerebrovascular disease. Front Aging Neurosci 2021; 13: 728007.

36.

Levy

Wagner

. Cognitive control and right ventrolateral prefrontal cortex: reflexive reorienting, motor inhibition, and action updating. Ann N Y Acad Sci 2011; 1224: 40–62.

37.

Liu

Duan

Zhang

, et al. Aberrant brain regional homogeneity and functional connectivity in middle-aged T2DM patients: a resting-state functional MRI study. Front Hum Neurosci 2016; 10: 490.

38.

Wang

Hou

Jiang

, et al. Alterations of spontaneous brain activity in type 2 diabetes mellitus without mild cognitive impairment: a resting-state functional magnetic resonance study. Front Hum Neurosci 2023; 17: 1305571.

39.

Ibrahim

Suppiah

Ibrahim

, et al. Diagnostic power of resting-state fMRI for detection of network connectivity in Alzheimer's disease and mild cognitive impairment: a systematic review. Hum Brain Mapp 2021; 42: 2941–2968.

40.

Cha

Kim

, et al. Functional alteration patterns of default mode networks: comparisons of normal aging, amnestic mild cognitive impairment and Alzheimer's disease. Eur J Neurosci 2013; 37: 1916–1924.

41.

Wang

Kong

, et al. Inversion of a large-scale circuit model reveals a cortical hierarchy in the dynamic resting human brain. Sci Adv 2019; 5: eaat7854.

42.

Liang

Tan

, et al. Altered functional hubs and connectivity in type 2 diabetes mellitus without mild cognitive impairment. Front Neurol 2020; 11: 1016.

43.

Michailidis

Moraitou

Tata

, et al. Alzheimer's disease as type 3 diabetes: common pathophysiological mechanisms between Alzheimer's disease and type 2 diabetes. Int J Mol Sci 2022; 23: 2687.

44.

Zhang

Tam

WWS

Hounsri

, et al. Effectiveness of combined aerobic and resistance exercise on cognition, metabolic health, physical function, and health-related quality of life in middle-aged and older adults with type 2 diabetes mellitus: a systematic review and meta-analysis. Arch Phys Med Rehabil 2024; 105: 1585–1599.