The Effectiveness of VR-Based Cognitive Training Program for Mild Cognitive Impairment: A Pilot Study

Abstract

Background and Purpose

Virtual reality (VR)-based cognitive training programs represent an emerging intervention for cognitive impairment. This pilot study aimed to evaluate the feasibility of a VR-based cognitive training program in patients with mild cognitive impairment (MCI).

Methods

Thirty-two patients diagnosed with MCI according to Peterson’s criteria participated in a 12-week VR training program, consisting of twice-weekly 50-minute sessions. Comprehensive assessments were conducted at baseline and after the intervention, including neuropsychological tests, questionnaires for depression, anxiety, quality of life, and dizziness severity. Caregivers evaluated patients’ daily living activities and neurobehavioral symptoms.

Results

Twenty-eight patients completed the program (87.5% women, mean age 73.21 years). Post-intervention analyses revealed significant improvements in both total composite and memory-specific scores on neuropsychological tests. No significant changes were observed in depression, anxiety, dizziness severity, neuropsychiatric symptoms, or daily living activities. Importantly, functional neuroimaging demonstrated statistically significant increases in connectivity among the bilateral hippocampus, parahippocampal gyrus, and amygdala, regions essential for memory and emotional processing.

Conclusions

This pilot study demonstrates that VR-based cognitive training may be a feasible therapeutic approach for cognitive impairment. The observed improvements in neuropsychological test scores and enhanced brain connectivity in memory-related regions suggest potential benefits for MCI patients. Further research with control group and larger sample sizes is warranted to confirm these findings and distinguish intervention effects from natural learning or test-retest effects.

INTRODUCTION

The aging population is experiencing a surge in the prevalence of cognitive impairments, including mild cognitive impairment (MCI) and dementia. These conditions not only impact the quality of life of affected individuals but also pose significant challenges for caregivers and healthcare systems.¹ To reduce the social and health burden of dementia, continuation of effort for prevention, early detection, and proper management are essential. MCI is the status that the patient shows objective cognitive impairment but preserved daily activity of life. Although recent studies have shown the efficacy of anti-amyloid treatments,² acetylcholinesterase inhibitors—a classical anti-dementia medication—have not demonstrated clinical efficacy in MCI status.³ In light of the limited efficacy and availability of pharmacological treatments for MCI, alternative therapeutic approaches are critically needed.

Cognitive training has emerged as a promising non-pharmacological intervention for cognitive impairments. The underlying premise is that targeted cognitive training can enhance the cognitive capacity and improve within the cognitive domain, potentially transferring to behavioral, neural, and biopsychosocial aspects.⁴,⁵ Furthermore, the baseline cognitive status, whether normal cognition or cognitive impairment, and magnitude of cognitive reserve may affect the outcomes of cognitive training.⁶,⁷ Numerous studies have explored various forms of cognitive training, with mixed results regarding their effectiveness.⁸,⁹,¹⁰

In this context, our study introduces a novel approach by incorporating virtual reality (VR) technology into cognitive training.¹¹ VR provides an interactive and virtual environment that can simulate real-life scenarios in both non-immersive and immersive ways, offering a unique platform for cognitive stimulation.¹² VR cognitive training provides advantages in terms of accessibility, cost-effectiveness, customization, and flexibility especially for the elderly, disabled individuals, those unable to frequently visit outpatient clinics.¹³ This pilot study, conducted across 3 medical centers in South Korea, aims to evaluate the efficacy and safety of a VR-based cognitive training program specifically designed for patients with MCI.

We hypothesize that compared to baseline status, the VR-based cognitive training program will lead to significant improvements in various cognitive domains, particularly memory, which is often one of the first areas affected in MCI.

METHODS

Subjects

This pilot study included 32 participants, who were part of the PREMIER (precision medicine platform for mild cognitive impairment on multi-omics, imaging, evidence-based R&D) study, a prospective, community-based cohort focusing on MCI. They visited the Dementia Center in Bucheon, South Korea, which is dedicated to comprehensive management of individuals with cognitive decline. The center offers services such as dementia prevention education, cognitive rehabilitation, early detection of cognitive impairment, and support for treatment. The participants underwent evaluation using the Korean version of clinical and neuropsychological (NP) assessment battery (Korean version of Consortium to Establish a Registry for Alzheimer's Disease [CERAD-K]), a detailed NP test.¹⁴ Expert neurologists analyzed all clinical data of the participants. The MCI criteria according to Petersen’s criteria¹⁵ included: 1) a memory complaint confirmed by an informant; 2) objective memory impairment relative to age, education level, and gender; 3) normal general cognitive function; 4) normal activities of daily living; and 5) absence of dementia. Objective memory loss was determined if the performance was 1.5 standard deviations below the norm for their specific age, education, and gender group in at least one of the 4 memory tests in the CERAD-K, which includes word list memory, word list recall, word list recognition, and constructional praxis recall tests. When the score of the Mini-Mental Status Examination falls within the normal range, adjusted for age and educational background, general cognitive function is considered to be normal. Regarding activities of daily living, when the Korean Instrumental Activities of Daily Living (K-IADL) was less than 0.40, they are deemed normal.¹⁶ Exclusion criteria include major psychiatric disorders, uncontrolled major medical conditions, contraindications for magnetic resonance imaging due to medical devices or claustrophobia, and a history of brain injury with loss of consciousness exceeding 30 minutes.

Intervention

The program was conducted individually using an immersive VR device connected to a tablet PC, a VR Head Mounted Display and Pico NEO 3 Eye (PICO, Beijing, China). The VR cognitive intervention consisted of a 12-week program, with sessions held twice per week (total 24 sessions). Each session included 50 minutes of VR-based training, focusing on cognitive stimulation for 30 minutes, physical exercise for 16 minutes, and meditation for 2 minutes. After cognitive training and physical exercise, participants take a break by removing the VR display.

All cognitive training tasks were originally developed for this study and have not been previously reported in prior literature. The cognitive stimulation program was designed to cover 6 cognitive domains: executive function, attention, visuospatial ability, calculation, memory, and language. Each domain contained 2 distinct task formats to ensure varied cognitive stimulation and minimize habituation over the 12-week intervention. All tasks were delivered in a gamified format and provided real-time audiovisual feedback to promote engagement and reinforce cognitive learning. The specific tasks for each domain are detailed in Table 1.

Table 1. Tasks of cognitive domains.

Cognitive domains	Task type	Description
Executive function	Rule-based tasks	Complete tasks based on assumed rules
	Sequenced response	Tap a tambourine of varying colors or notes as instructed
Attention	Visual discrimination	Identify spinning, overlapped numbers
	Response to stimulus	Play a xylophone according to presented colors or notes
Visuospatial ability	Object counting	Count the number of circling boxes
	Spatial puzzle	Complete empty puzzles with correct spatial arrangements
Calculation	Mental arithmetic	Identify numbers on dice and perform calculations
	Price calculation	Sum the total price of items in a grocery shopping scenario
Memory	Object location recall	Recall the locations of various items after presentation
	Biographical information	Remember and recall information about presented individuals
Language	Word search	Locate words arranged horizontally or vertically in mixed characters
	Word completion	Complete missing syllables in partially presented words

All tasks were designed with 3 difficulty levels (easy, moderate, hard), and the system adaptively adjusted progression based on individual performance to maintain optimal challenge.

For executive function, tasks were designed to target rule inference, inhibitory control, and cognitive flexibility. Patients complete tasks based on assumed rules and tap a tambourine of varying colors or notes as instructed. In the attention domain, tasks aimed to enhance sustained attention and working memory. Participants identify spinning, overlapped numbers and play a xylophone according to presented colors or notes. For the visuospatial domain, patients count the number of circling boxes and complete empty puzzles, stimulating visual tracking and spatial reasoning. In the calculation, they perform mental arithmetic to identify numbers on dice and sum the total price of items in a grocery scenario. For memory domain, patients recall the locations of various items and information about presented individuals, engaging both visual and episodic memory processes. In the language domain, patients locate words arranged horizontally or vertically in a panel of mixed characters and complete missing syllables in words, stimulating lexical retrieval and orthographic processing. The challenging levels included easy, moderate, and hard modes, where successful completion of one difficulty level led to an escalation in challenge for the subsequent trial, consistent with evidence suggesting that adaptive difficulty enhances cognitive training outcomes.¹⁷ The adaptive difficulty algorithm automatically adjusted task complexity based on individual performance, providing a personalized and progressively challenging training experience.

Physical exercise program offers the training whether during sitting or standing. The physical exercise program offers training adaptable for execution in either a seated or standing position. Participants opting for standing exercises engage in marching in place, rotating clockwise and counterclockwise, waving one arm followed by both arms, and drawing circles with each arm. Those selecting seated exercises perform arm waves, trunk leans in various directions, arm circles, and drum hitting. During the meditation period, participants practice deep breathing against a backdrop of natural scenery and birdsong.

Assessment

For assessing the outcomes, we utilized NP tests and questionnaires both at the baseline and at the end of the 12-week program. For assessment of cognitive function, participants underwent a computerized NP test (Inbrain Cognitive Screening Test [CST]),¹⁸ which comprises 7 NP assessments targeting 5 cognitive domains: forward and backward tasks in the Visual Span Test for attention; the Difficult Naming Test and word fluency tests for both semantic (fruits) and phonemic (Korean alphabet ‘digeut’) categories for language domain; the Block Design Test for visuospatial domain; immediate recall, delayed recall, and recognition of Word Place Association Test (WPAT) for memory; and the Korean-Trail Making Test-Elderly version for executive functions. The Inbrain CST provides a total score and subscores within 5 cognitive domains. The subscores for the 5 domains and the overall composite score of the Inbrain CST are standardized on a scale from 0 to 100. The total score is calculated from the aggregation of subscores, each assigned different weights: attention (0–20), language (0–23), visuospatial function (0–10), memory (0–26), and executive function (0–21). Higher scores indicate superior cognitive performance.

Initially, every participant responded to various questionnaires covering depression (short form of the geriatric depression scale, with 15 items scored 0–15; higher scores indicate more severe symptoms),¹⁹ anxiety (geriatric anxiety inventory, with 20 items scored 0–20; higher scores indicate more severe symptoms),²⁰ and quality of life (Geriatric Quality of Life-Dementia, with 15 items scored 15–60; higher scores indicate a better quality of life).²¹ To assess safety and discomfort related to intervention, participants completed questionnaires regarding the intensity of dizziness (University of California Los Angeles Dizziness Questionnaire [UCLA-DQ], with 5 items scored 5–25; higher scores indicate more severe symptoms)²² and the severity of simulator sickness (Simulator Sickness Questionnaire [SSQ], which includes 16 items yielding scores for nausea, oculomotor, disorientation subscales, and a total score, with higher scores indicating more severe symptoms).²³ Caregivers also provided evaluations of the participants’ capabilities in K-IADL¹⁶ and their neurobehavioral symptoms (neuropsychiatric inventory).²⁴ These evaluations and questionnaires were administered once more after the program sessions were completed.

The primary outcome was the total score of Inbrain CST. Secondary cognitive outcomes included domain-specific composite scores from Inbrain CST and individual memory test scores (immediate, delayed recall, and recognition of WPAT).

Imaging acquisition and functional connectivity analysis

Participants in the study underwent resting-state functional magnetic resonance imaging (fMRI) scans at 2 time points: baseline and 12 weeks post-participation in the VR-based training program. The scans were performed using a 3T Magnetom Skyra MRI system (Siemens Healthineers, Erlangen, Germany) located at Bucheon Soonchunhyang Hospital. During the scanning session, participants were asked to lie down, close their eyes, and rest. Whole-brain functional images were obtained using a fast field-echo planar imaging sequence (repetition time=3,000 ms, echo time=30 ms, matrix=64×64, field of view=224 mm, flip angle=90°, slice thickness=3.5 mm/0 gap, slices=42, volumes=120, total time=369 seconds).

The CONN toolbox (release 20b),²⁵ running on Statistical Parametric Mapping version 12 (University College London/Queen Square Institute of Neurology, London, UK) and Matlab 2016a (MathWorks, Natick, MA, USA), was used for functional connectivity analysis. Image preprocessing involved 5 standard steps in CONN: realignment, slice timing correction, segmentation and normalization to Montreal Neurological Institute space, scrubbing using artifact detection tools (global signal threshold >5 z value, participant motion threshold >0.9 mm), and smoothing with a 6-mm full width at half maximum Gaussian kernel. After preprocessing, the time series underwent band-pass filtering (0.008–0.09 Hz) and regression of white matter and cerebrospinal fluid time series.

The Harvard-Oxford cortical and subcortical atlas was used to parcellate the brain into 132 regions of interest (ROIs).²⁶ Whole-brain functional connectivity analysis (ROI-ROI) was conducted following the methodology described in our earlier works, with only minor modifications.²⁷,²⁸,²⁹ For each ROI, the average time-series was derived from the preprocessed images. Pearson correlation coefficients, serving as a measure of functional connectivity, were computed for all pairs of the 132 time-series and converted to -scores using Fisher’s r-to-z transformation. Lastly, whole-brain functional connectivity was assessed for changes between the pre- and post-cognitive training time points.

To assess alterations in functional connectivity after cognitive training, we used the threshold-free cluster enhancement (TFCE) method³⁰ for multiple comparison correction. As a first step, we utilized the hierarchical optimal leaf ordering algorithm,³¹ as implemented in the CONN toolbox, to sort a connectivity map encompassing 132 ROIs and their corresponding ROI-ROI pairs. Next, we applied the CONN toolbox’s default statistical settings for TFCE analysis to identify significant clusters of ROI-ROI connectivity that demonstrated an increase in connectivity following cognitive rehabilitation. This process generated a TFCE score for each identified cluster. To establish the statistical significance of these TFCE scores, we used a family-wise error (FWE)-corrected threshold of p<0.05, which was derived by performing 1,000 permutations of the data.

Statistical analysis

For continuous variables, the Wilcoxon signed-rank test was adopted to compare scores from the NP battery and various questionnaires between baseline and after the 12-week (24 sessions) intervention. The SSQ scores were averaged over the first week (encompassing the first and second sessions) and the twelfth week (covering the twenty-third and twenty-fourth sessions). The Mann-Whitney test was used to compare the average SSQ scores between the first and twelfth weeks.

RESULTS

Out of the 32 enrolled patients with MCI, 28 successfully completed the 12-week VR-based cognitive training program. Four participants withdrew from the study: one due to poor vision, which hindered the administration of the VR cognitive training, and 3 owing to personal reasons, including discomfort with the frequency of visits. The majority of these participants were women, constituting 24 out of 28 (87.5%), with an average age of 73.21±4.20 years. The average education level was 8.61±4.55 years. The primary outcome, assessed through the Inbrain CST NP test, revealed significant improvements post-training. Specifically, the total composite score exhibited a noteworthy increase from the baseline score of 49.04±9.15 to 52.96±9.53 post-intervention (p<0.0001). Similarly, the memory-composite score showed a substantial improvement, rising from 51.36±35.98 at baseline to 66.44±31.40 after the 12-week training period (p=0.0037) (Table 2, Fig. 1).

Table 2. Results of neuropsychological tests.

Variables		Baseline	Follow-up	p-value
MMSE		26.43±2.03	27.25±2.47	0.0947
CST
	Total composite score	49.04±9.15	52.96±9.53	<0.0001*
	Attention subscore	62.52±32.16	73.10±23.97	0.0735
	Language subscore	56.58±29.78	65.58±28.41	0.1200
	Visuospatial subscore	60.31±31.16	64.40±30.21	0.0934
	Memory subscore	51.36±35.98	66.44±31.40	0.0037*
	Frontal/Executive subscore	65.58±26.83	69.98±23.23	0.4117

Data are expressed as the mean with standard deviation.

MMSE: Mini-Mental Status Examination, CST: computerized screening test.

^*p<0.05.

Fig. 1. Changes in composite and sub-domain scores. Compared to the baseline, it showed significant improvement in the (A) total composite score and (E) memory function. (B) Attention, (C) language, (D) visuospatial, and (F) executive functions were not significantly different between the baseline and 12-weeks FU.

FU: follow-up.

^*p<0.05.

In terms of secondary outcomes encompassing mental health and daily functioning, the study did not observe significant changes between baseline and post-training evaluations. These measures included depression, anxiety, neuropsychiatric symptoms, and the ability to perform instrumental activities of daily living, assessed through various questionnaires and scales. Regarding safety and discomfort related to intervention, there was no significant difference in the UCLA-DQ scores before and after the intervention (p>0.05). The nausea, oculomotor, disorientation subscales, and the total score of the SSQ demonstrated a significant decrease in the twelfth week compared to the first week (Table 3).

Table 3. Comparative analysis of psychiatric symptoms, quality of life, and daily functioning pre- and post-intervention.

Variables		Baseline	Follow-up	p-value
K-IADL		0.059±0.113	0.041±0.151	0.1567
SF-GDS		5.759±1.683	5.241±1.300	0.1727
GAI		4.655±4.546	3.793±5.551	0.1042
GQOL-D		35.68±7.991	34.25±8.414	0.2552
NPI		1.929±4.198	1.786±4.158	0.9521
UCLA-DQ		8.036±4.150	7.250±3.503	0.2275
SSQ
	Nausea	3.066±5.228	0.000±0.000	0.0003*
	Oculomotor	4.196±7.022	0.000±0.000	0.0001*
	Disorientation	5.966±9.028	0.000±0.000	0.0003*
	Total score	4.875±6.384	0.000±0.000	<0.0001*

Data are expressed as the mean with standard deviation.

K-IADL: Korean Instrumental Activities of Daily Living, SF-GDS: short form of the geriatric depression scale, GAI: geriatric anxiety inventory, GQOL-D: Geriatric Quality of Life-Dementia, NPI: neuropsychiatric inventory, UCLA-DQ: University of California Los Angeles Dizziness Questionnaire, SSQ: Simulator Sickness Questionnaire.

^*p<0.05.

Functional connectivity analysis revealed a significant cluster of increased connectivity following cognitive training (TFCE=95.77, p-FWE=0.032; Table 4). The cluster primarily encompassed core regions involved in episodic memory processing, including the bilateral hippocampus, parahippocampal gyrus, posterior cerebellum, and amygdala (Fig. 2).

Table 4. Increased functional connectivity after cognitive training.

Functional connectivity (ROI-ROI)		T-value
pPaHC R
	Hippocampus R	4.84
	Hippocampus L	5.83
	Amygdala R	3.01
	Amygdala L	1.88
	pPaHC L	4.47
pPaHC L
	Hippocampus R	1.73
	Hippocampus L	1.82
Amygdala L
	Cereb9 R	1.80

Threshold free cluster enhancement, family-wise error corrected threshold of p<0.05.

ROI: region of interest, pPaHC: posterior division of the parahippocampal gyrus, R: right, L: left, Cereb9: cerebellum 9.

Fig. 2. Changes in functional connectivity. (A) Increased functional connectivity post-cognitive training. This set of brain images illustrates the increased functional connectivity in the bilateral hippocampus, parahippocampal gyrus, and amygdala when compared to the baseline. (B) Connectivity matrix. The color-coded matrix displays the enhanced functional connectivity between the regions of interest: the left and right amygdala, hippocampus, parahippocampal gyrus (aPaHC and pPaHC), and the cerebellum.

aPaHC: anterior division of the parahippocampal gyrus, pPaHC: posterior division of the parahippocampal gyrus, L: left, R: right, Cereb9: cerebellum 9.

DISCUSSION

The findings of our study indicate that a 12-week VR-based cognitive training program can feasibly be implemented in patients with MCI, with potential improvements in cognitive functions, particularly memory. This is evident from the notable improvements in the total and memory-specific composite scores of the Inbrain CST NP test.

Until now, effective treatments for MCI status have been limited. Various nonpharmacological interventions, including physical exercise, cognitive training, and diet, have been investigated and shown advantages, but have failed to prevent dementia progression.³² Recently, it has been recognized that multidomain intervention is more effective than single-domain intervention, and studies such as FINGER,³³ SUPERBRAIN,³⁴ and U.S. POINTER³⁵ are actively under investigation. The multidomain intervention, including modification of diet, vascular risk factor monitoring, and cognitive training, can be administered in hospital settings, outpatient clinics, or at home with technological support. Moreover, with the occurrence of the pandemic, an effective and easily accessible cognitive training method was needed.

Meta-analysis on the effectiveness of cognitive training in MCI found that compared to control interventions, there were slight to moderate beneficial effects on global cognitive function, memory, language, working memory, and executive function.³⁶,³⁷ Although the duration, frequency, and type of training were variable across the studies, these factors little affect the effect size of the cognitive training efficacy.³⁷ In comparison to restorative and compensatory cognitive training approaches, multicomponent cognitive training demonstrated a notably significant and moderate effect size.³⁷ In this study, the VR program consists of multicomponent cognitive training, targeting various cognitive domains, including memory, language, visuospatial abilities, executive functions, attention, and calculation. The distinguishing feature of this intervention is its adaptive design, which automatically adjusts task complexity based on each participant’s performance. This personalization is especially important for individuals with MCI, who often show considerable inter-individual variability in cognitive functioning. Setting an appropriate intensity to stimulate cognitive function could potentially enhance cognitive abilities more efficiently.¹⁷ By continuously calibrating the level of challenge, the program promotes engagement and maintains motivation throughout the intervention. This adaptive approach likely contributed to the high adherence observed in the study, with 87.5% of participants completing the full 12-week program. Considering the participants’ advanced age and cognitive impairment, such a completion rate highlights the feasibility and acceptability of the intervention in real-world clinical settings.

Compared to simple computer-based cognitive training, the VR cognitive stimulation program provides a much closer approximation to real-life environments.³⁸,³⁹,⁴⁰ The immersive nature of VR, offering a rich, engaging, and controlled environment, may provide a more effective stimulus for cognitive training than traditional methods.⁴⁰ These findings contribute to the understanding of VR as a tool in cognitive training.³⁸ A meta-analysis with computer-based cognitive training and VR assisted training, the latter showed greater effect size in most of the cognitive domains.³⁸ While the introduction of VR technology has raised safety concerns regarding VR sickness, particularly among older adults, most participants with dementia tolerated immersive VR interventions well.⁴¹ In this study, tolerability was assessed by the SSQ. Although mild discomfort was reported during the first week (mean SSQ scores <10, indicating negligible [<5] or minimal [5–10] levels), all scores decreased over time and averaged zero by the final week (Table 3). These results suggest high tolerability and rapid adaptation to the immersive VR environment, which is particularly important in older adults who are more vulnerable to dizziness and fall risks. Furthermore, because the VR training system requires only a commercially available headset and tablet-based interface, it can be implemented outside of traditional clinical settings. This flexibility supports broader accessibility, especially for patients with mobility limitations or those who require caregiver assistance to attend in-person sessions.

The outcome of cognitive function can be assessed through task performance in specific cognitive domains, as well as questionnaires evaluating IADL, neuropsychiatric behaviours, and psychiatric aspects. Furthermore, objective assessments using neuroimaging techniques such as brain volume and functional connectivity can be employed to measure changes resulting from cognitive training.⁴² With resting-state fMRI, randomized controlled trials with computerized cognitive training in MCI⁴³ and healthy elderly⁴⁴ showed increased connectivity between the hippocampus and the left superior frontal lobe. The increased brain connectivity correlated with changes in memory function among MCI⁴³ and global cognition among healthy elderly.⁴⁴ The hippocampus and parahippocampus are core regions for episodic memory, visuospatial processing, navigation, and general cognitive processing.⁴⁵,⁴⁶ And these areas are the most vulnerable regions for Alzheimer’s disease.⁴⁷ In our study, we examined altered brain connectivity following VR cognitive training, revealing increased brain activity between the parahippocampus and hippocampus, parahippocampus and amygdala, and amygdala and cerebellar hemisphere. Moreover, compared to baseline, participants showed better performance in the total composite score and memory composite subscore, which are related to the hippocampus and parahippocampus. It remains unclear whether the observed changes in functional connectivity directly correspond to cognitive or clinical improvements, although they were observed in memory-related brain regions. Functional connectivity metrics provide only indirect evidence of neuroplasticity and should therefore be interpreted with caution.

Interestingly, our study showed altered brain connectivity with the amygdala and cerebellar 9 area. Considering the amygdala’s significant roles in attention, associative learning, and emotion processing,⁴⁸,⁴⁹ this multicomponent VR cognitive training may affect the functional connectivity of the amygdala. To respond appropriately following the instructions, it may be necessary to remain alert and vigilant during the session. The cerebellum is an important structure involved in tuning sensorimotor processing. Emotional and cognitive processing also occur in the cerebellum, particularly in its posterior lobe.⁵⁰ The lobule IX is primarily connected with associative cognitive and limbic regions.⁵¹ Our findings also revealed increased brain connectivity between cerebellar lobule IX and amygdala areas.

However, it is important to note that this enhancement in cognitive performance did not extend to other aspects of mental health and daily functioning, such as depression, anxiety, and neuropsychiatric symptoms. The lack of significant change in these secondary outcomes suggests that while VR-based cognitive training is effective in targeting specific cognitive domains, its impact on broader mental health and daily functioning may be limited. This finding aligns with the growing body of research indicating that cognitive training can yield domain-specific improvements.⁸,⁵² It also underscores the complexity of MCI and the multifactorial nature of its impact on patients’ lives.

Taken together, the findings from this pilot study suggest that VR-based cognitive training is a feasible and potentially beneficial intervention for MCI. However, several limitations inherent in the study design warrant cautious interpretation of the results. First, this study is a single-arm pilot study without a control group, the findings are vulnerable to confounding factors such as natural learning effects or test-retest effects. Second, although significant changes were observed in cognitive performance and resting-state functional connectivity, the absence of a control group limits causal inferences regarding the effects of the intervention. Third, the durability of the cognitive improvements observed after the 12-week intervention remains unclear. Future studies incorporating long-term follow-up assessments will be necessary to determine whether sustained cognitive benefits can be achieved without ongoing intervention. Fourth, although the rs-fMRI analysis provided insights into neuroplastic changes, establishing causality between functional connectivity and cognitive improvement requires more sophisticated imaging analyses and larger samples. Lastly, the small sample size also limits the generalizability of the results. Future research should aim to include larger and more diverse samples and possibly adopt a randomized controlled trial design to strengthen the evidence for the effectiveness of VR-based cognitive training in MCI.

In conclusion, our pilot study demonstrated post-VR cognitive training changes in cognitive functioning compared to baseline, alongside confirmed alterations in brain connectivity among patients with MCI. It also opens avenues for further research to explore the potential of VR technology in cognitive rehabilitation, particularly in neurodegenerative conditions where treatment options are limited.