Preliminary study on the feasibility of virtual reality-based cognitive training on patients with mild to moderate Alzheimer's disease

Abstract

Background

Alzheimer's disease (AD) is a neurodegenerative disease that causes a decline in cognitive functions, considerably affecting a patient's life. Recently, virtual reality (VR) technology has emerged as a new tool used in the cognitive training of patients with AD.

Objective

This study aimed to investigate the safety, feasibility, and clinical efficacy of VR-based cognitive training for patients with mild to moderate AD.

Methods

Thirteen participants diagnosed with mild to moderate AD underwent VR training sessions by using the MentiTree software. Each session was conducted for 30 min twice a week for 9 weeks (total of 540 min). Cognitive functions were assessed before and after the intervention.

Results

Although 1 of the 13 participants experienced adverse effects, the 9-week cognitive training was well tolerated and had a high feasibility of 93%±24.65%. A tendency toward improvement was observed in the visual recognition memory of the participants (p = 0.034), but other domains did not significantly change.

Conclusions

VR-based cognitive training is safely accepted by patients with mild to moderate AD. The potential of VR in AD treatment should be further explored using a randomized control group.

Introduction

Alzheimer's disease (AD) is the most common type of dementia that manifests as a decline in the ability to perform daily tasks and decline in cognitive functions, such as memory, language ability, and judgment; it is accompanied by behavioral psychological symptoms that impair the quality of life of patients and their caregivers. Its exact cause has not yet been identified, but the most widely accepted cause is based on the amyloid hypothesis, which states that AD is attributed to the accumulation of amyloid-β and tau proteins. ¹

The incidence of AD is correlated with age. ² In Korea, the incidence and prevalence of AD in adults aged 40 years or older increase yearly as the population ages. ³ Currently, AD is commonly treated using symptomatic treatments, but they only alleviate the symptoms of AD and do not provide a fundamental solution. ⁴ Recently approved disease modifying drugs require large doses because they must pass through the blood–brain barrier, potentially increasing the risk of side effects. ⁴ For these reasons, relatively non-invasive and safe cognitive training and cognitive rehabilitation that can be combined with pharmacological treatment have been widely considered.^5,6 Regular and continuous cognitive training (CT) enhances cognitive reserve, ⁷ which means resilience to neuropathological damage of the brain, and promotes neuroplasticity in relevant brain networks. ⁸ In particular, cognitive training using digital devices, such as computerized cognitive training (CCT) or virtual reality cognitive training (VRCT) offers a more cost-effective, accessible, flexible, and comprehensive intervention. ⁹ These approaches can also provide individualized intervention such as increasing task difficulty incrementally to ensure that the intervention remains sufficiently challenging. This flexibility can promote the potential for plasticity-induced changes to occur. ¹⁰ Recent studies have shown that incorporating mobile applications into cognitive training may provide several additional benefits, including improved cognitive functions and enhanced accessibilities. ¹¹

Among various ways to improve cognitive training, virtual reality (VR), which involves computer technology to create a virtual 3D environment where users can move and interact in real time, ¹² is gaining attention. It can provide users with a high sense of immersion and a personalized experience similar to their real environment.^9,13 Studies have shown that VR technology can improve the cognitive function of patients with AD.^14–16 However, it is often accompanied by cybersickness, which is similar to motion sickness such as eye fatigue, headache, and nausea.^17,18 Furthermore, it cannot be easily used by elderly people or patients who have cognitive impairment and are not familiar with operating electronic devices or virtual spaces.

Therefore, the adverse effects of VR should be reduced to use VR more widely in AD treatment, and an intuitive and easy-to-use interface should be developed.^19,20 In the present study, we aimed to determine whether patients with mild to moderate AD could tolerate VR-based cognitive training by using the MentiTree software without experiencing side effects; we also aimed to elucidate the clinical efficacy of the program.

Methods

Participants

The subjects were adults aged 50–90 years and diagnosed with mild to moderate AD. The inclusion criteria were as follows: (1) subjects with a history of more than 24 weeks with the National Institute on Aging-Alzheimer's Association clinical stage ²¹ ≥4 as of the screening date and a Clinical Dementia Rating (CDR) Global score of 0.5–2, (2) subjects receiving a drug or non-drug treatment to improve or alleviate the cognitive function of AD, and (3) subjects receiving a stable treatment for more than 12 weeks as of the screening date.

Patients with the following conditions were excluded: vascular dementia, cerebrovascular disease, heart failure, body mass index >40, uncontrolled hypertension or diabetes, malignant tumor not cured within 5 years, comorbid psychiatric diagnosis and currently experiencing significant symptoms, infectious or metabolic diseases that may cause cognitive decline, severe or unstable physical illness, inability to perform basic psychomotor activities that cause difficulty in assessing the questionnaire, or hearing or visual impairment that hampers the use of clinical trial medical devices and/or general equipment (VR).

The purpose and method of the study and the protection of the rights of research participants were approved by the Bioethics Review Committee of Kyungpook National University Chilgok Hospital (Approval no: KNUCH 2022-10-027-002). All participants were fully informed of the experiment and enrolled as experimental subjects after voluntary consent was obtained from them or their guardians.

Intervention

The participants underwent VR training using MentiTree^22,23 twice a week for 30 min at a time for 9 weeks. In the MentiTree program, they wore a head-mounted display (HMD, Oculus Rift S, 2560 × 1440 resolution, 115-degree field of view) and received a game-style 3D VR content to perform cognitive function training. They were able to interact with objects in the VR by using a hand tracking technology,^24,25 which recognizes the user's hand movements through sensors and projects them onto virtual hand movements on the screen. MentiTree alternately provides an indoor background content (randomly provided) and an outdoor background content (sequentially provided); its difficulty level (indoor background content: 1–5 levels; outdoor background content: 1–3 levels) automatically changes depending on the subject's ability to perform the content. The indoor background content included the following: (1) making a sandwich, (2) using the bathroom, (3) tidying up the playroom, (4) choosing a book, (5) making cereal, (6) playing a cartoon, (7) matching blocks, (8) gathering supplies, (9) dressing a doll, (10) making a phone call, (11) playing a game, (12) solving card problems, (13) playing at the market, and (14) coloring. Figure 1 shows an example of indoor VR training tasks. The outdoor background content covered the following: (1) finding directions, (2) guessing friends, (3) finding Yeong-hee, (4) guessing the number of balloons, (5) guessing animals, (6) guessing the order of the cards, (7) guessing the phone number, (8) pressing the vending machine, (9) guessing the answer to a song, (10) shopping, and (11) finding the way home. Figure 2 shows an example of outdoor VR training tasks.

Figure 1.

An example of the tasks used in MentiTree training program (indoor content).

Figure 2.

An example of the tasks used in MentiTree training program (outdoor content).

Neuropsychological assessment

The cognitive function of the participants was assessed using the Korean version of the Mini-Mental State Examination-2 (K-MMSE-2), ²⁶ Clinical Dementia Rating (CDR), ²⁷ Global Deterioration Scale (GDS), ²⁸ and Literacy Independent Cognitive Assessment (LICA) ²⁹ before and after the intervention. LICA consists of the following 13 subtests: ³⁰ (1) story recall test-immediate recall, (2) stick construction test, (3) word recall test-immediate recall, 4)forward/backward visuospatial span test, (5) digit Stroop test, (6) calculation test, (7) story recall test-delayed recall, (8) story recall test-recognition, (9) stick recognition test, 10) word recall test-delayed recall, (11) word recall test-recognition, (12) word fluency test-animal name quotient, and (13) color and object recognition test and naming.

LICA is a cognitive function assessment tool that reflects the characteristics of illiteracy and can be applied to illiterate and low-educated elderly people. ³⁰ It has high sensitivity and specificity for dementia diagnosis ²⁹ and is known to indicate the severity of dementia efficiently. ³¹

The participants in this study were classified as having mild to moderate dementia; because of the characteristics of the subjects and their residential areas, they tended to be elderly or have relatively low levels of education and cognitive test performance,^32,33 which may limit their participation. ²⁹ They might also reluctantly perform high-difficulty tests or experience difficulty in distinguishing performance levels. ³⁴ As such, LICA was implemented considering the characteristics of these participants and the local community.

Statistical analysis

Research data were analyzed using the SciPy library in Python. The normality of the change in scores before and after each test was examined using the Shapiro–Wilk test. When normality was met (normal distribution), a paired sample test was used; otherwise (nonparametric distribution), the Wilcoxon signed-rank test was conducted to compare the test scores before and after the intervention. The test scores were expressed as mean ± standard deviation, and their significance level was set at 0.05. The Bonferroni correction was applied to adjust for multiple comparisons.

Results

A total of 13 people participated in the experiment and received VR training at least once. One of them dropped out, and the remaining patients completed the test. The sociodemographic variables and pre-intervention K-MMSE-2, CDR, and GDS of the 12 patients who completed the experiment are shown in Table 1.

Table 1.

General characteristics of the subjects.

Variable	Mean ± SD	N (%)
Demographic Data
Female	N/A	10 (83.3)
Male	N/A	2(16.7)
Age (y)	74 ± 10.56 (55–87)	N/A
Education (y)	6.83 ± 4.33 (0–15)	N/A
Pre-Intervention Neuropsychological Assessment
K-MMSE-2	17.25 ± 6.11	N/A
CDR	1.17 ± 0.54	N/A
GDS	4.42 ± 0.79	N/A

Values are means ± standard deviation (SD). K-MMSE-2: Korean version of Mini-Mental State Examination, 2nd edition; CDR: Clinical Dementia Rating Scale; GDS: Global Deterioration Scale.

The adherence rate of training was expressed as a percentage of how many of the 18 training sessions participants participated in. Twelve of the 13 participants completed all 18 training sessions; one participant attended only two sessions. Therefore, the mean adherence rate to the VR intervention was 93% (SD = 24.65). However, as training gains per session were not specifically recorded, we cannot determine the extent to which different participants fully participated in the training intervention. Among the 13 participants who received intervention more than once, only one experienced dizziness during training. However, before consenting to participate, this participant had a history of dizziness and earache for several years. Nonetheless, vertigo is one of the well-known types of cyber sickness ³⁵ and among the various factors considered to cause cyber sickness, visual-vestibular conflict is assumed to be similar to motion sickness. ³⁶ Therefore, despite recent technological efforts to manage cyber sickness through various devices or contents, more caution should be taken when applying it to subjects with a history of vestibular -related disorders or vestibular hypersensitivies. ³⁷ Except for this patient, all patients had no adverse effects. Therefore, the experiment was well tolerated.

Table 2 shows the mean ± standard deviation of each test result before and after the intervention. The scores of the LICA stick recognition test, which measures the visual recognition memory of patients before and after the VR intervention, increased statistically significantly (p = 0.034). However, the results of the other tests did not significantly differ between the pre- and post-intervention groups. After Bonferroni adjustment (adjusted p <0.003), none of the result remained significant.

Table 2.

Comparisons of neuropsychological assessment results before and after 9 weeks of VR training.

Variable	Baseline	9 weeks	Test type	Effect size [95%CI]	p
K-MMSE-2	17.25 ± 6.11	16.92 ± 6.04	Paired t-test	d = −0.20 [−1.39,0.73]	0.4333
CDR	1.17 ± 0.51	1.25 ± 0.56	Wilcoxon	r = −1.00 [−1.00,−1.00]	0.3173
GDS	4.42 ± 0.79	4.58 ± 0.79	Wilcoxon	r = −1.34 [−0.95,−0.95]	0.1573
Total LICA	138.09 ± 38.03	141.11 ± 49.01	Wilcoxon	r = −1.26 [−0.82,−0.02]	0.2334
Story immediate recall	4.08 ± 2.01	3.08 ± 2.02	Wilcoxon	r = −1.75 [−0.89,−0.07]	0.0782
Story delayed recall	0.25 ± 0.89	0.33 ± 0.89	Wilcoxon	r = −1.00 [−1.00,−1.00]	0.3173
Story recognition	4 ± 1.91	3.75 ± 2.05	Paired t-test	d = −0.13 [−1.47,0.97]	0.5839
Stick construction	7.17 ± 3.66	6.75 ± 4.02	Wilcoxon	r = −0.18 [−0.64,−0.03]	0.7150
Stick recognition	11.58 ± 3.29	14.00 ± 3.30	Paired t-test	d = 0.64 [0.08,1.74]	0.0342
Word immediate recall	9.83 ± 4.20	10.33 ± 4.89	Paired t-test	d = 0.14 [−1.47,0.97]	0.6881
Word delayed recall	0.25 ± 0.87	0.25 ± 0.62	Wilcoxon	r = 0.00 [−0.95,0.00]	1.00
Word recognition	12.50 ± 3.15	12.83 ± 3.81	Paired t-test	d = 0.12 [−1.39,2.05]	0.6338
VSS forward	3.67 ± 2.02	3.33 ± 2.02	Paired t-test	d = −0.34 [−0.96,0.29]	0.2342
VSS backward	2.33 ± 1.67	2.83 ± 1.70	Paired t-test	d = −0.40 [−1.29,0.29]	0.1962
Digit Stroop test	12.17 ± 8.74	11.83 ± 8.03	Paired t-test	d = −0.06 [−3.85,3.18]	0.8228
Calculation	14.75 ± 8.43	15.25 ± 9.10	Paired t-test	d = 0.17 [−1.35,2.35]	0.6746
Animal fluency	7.08 ± 3.90	7.75 ± 5.67	Paired t-test	d = 0.21 [−1.40,2.74]	0.7202
CORT	10.42 ± 2.75	10.92 ± 3.87	Paired t-test	d = 0.23 [−0.87,1.87]	0.4924
Naming	11.00 ± 3.36	10.83 ± 3.59	Paired t-test	d = −0.13 [−1.02,0.68]	0.6707

Values are means ± standard deviation (SD). CI: Confidence intervals; K-MMSE-2: Korean version of Mini-Mental State Examination, 2nd edition; CDR: Clinical Dementia Rating Scale; GDS: Global Deterioration Scale; LICA: Literacy Independent Cognitive Assessment; VSS: Visuospatial Span; CORT: Color and Object Recognition Test.

MentiTree automatically adjusts the task difficulty by increasing, maintaining, or decreasing the training level based on the user's previous performance on the corresponding mission. All participants started at Level 1 on the first day and changed their level according to their previous training performance. For the 12 participants who completed all 18 training sessions, all of them showed an increase in difficulty level on the final session compared to the first session, and the average difficulty level increased from an initial level of 1 to a final level of 2.85 for outdoor contents, and from 1 to 3.9 for indoor content.

Discussion

This study aimed to investigate the feasibility and changes in the cognitive function of patients with mild to moderate AD after 9 weeks of cognitive training using the MentiTree VR program. The safety of the program was high, and no adverse effects other than a single episode of dizziness were observed. The average adherence rate to the training was also high, at 93% (SD = 24.65). While the stick recognition test showed statistically significant improvement before correction, it did not reach the adjusted threshold for multiple comparisons. This finding may suggest a trend toward improvement in visual recognition memory, although it should be interpreted cautiously. Given the small sample size and exploratory nature of this pilot study, further validation with a larger cohort is warranted to determine whether this effect is robust and clinically meaningful.

Studies have been using the HMD equipment to create a VR content for cognitive training. In the MentiTree program used in the present study, the users wore the HMD and interacted with VR by using their hands through the hand-tracking technology. Thus, their intuitiveness was enhanced, allowing the patients with AD and cognitive decline to easily use the program; their immersion was also improved. ²⁴ In addition, through training using the hands, the patients could exercise their fine motor skills by inducing hand movements such as grasping and stretching.

The MentiTree program provides a scenario-based training. Each training is connected to a single context of performing a mission with a granddaughter on the topic of daily life. It adjusts the difficulty level according to the individual's performance and provides immediate feedback to the user through the granddaughter character. Therefore, it can provide greater user enjoyment, self-efficacy, and motivation. ³⁸ In addition, the MentiTree removes unreal VR elements, such as bringing distant objects, to increase realism. The high realism level obtained through MentiTree helps develop a naturalistic behavior, thereby increasing ecological validity.^39,40 VR has an additional advantage of convenience in collecting data on patients’ behavioral variables and physical movements. ³⁹

In the present study, visual recognition memory showed an increasing trend. Other studies have also shown that recognition memory is closely related to the medial temporal lobe, which includes the hippocampus and perirhinal cortex,^41,42 which are known as the most vulnerable regions for Alzheimer's disease. ⁴³ A previous study showed that a hand movement exercise program targeting elderly people with dementia can effectively enhance the cognitive function of participants. ⁴⁴ In addition, virtual environmental enrichment provided by 3D video games improves hippocampus-related memories, including recognition memory. ⁴⁵ Also, other recent study revealed that after VR cognitive training, there was increased brain activity between the parahippocampus and hippocampus, and improvement in memory composition score. ⁴⁶ Therefore MentiTree's fine muscle exercise using the hand together with providing various stimuli through VR, possibly contributed to the improvement in visual recognition memory. In patients with AD, impaired visual recognition memory can cause confusion and a sense of disconnection by causing difficulties in recognizing familiar environments and people. The possibility of amnestic MCI progressing to probable AD can be predicted through the decline in visual recognition memory.^47,48 Therefore, improving visual recognition memory can positively influence the treatment and prevention of AD. However, this study did not elucidate the mechanism by which the VR program improves the patients’ visual recognition memory; therefore, further studies should use fMRI to reveal the underlying mechanism.

The difficulty of the Mentitree program changes depending on participants’ performance; therefore, an increase in difficulty level indicates an improvement in task performance. In this training, participants’ final difficulty levels increased compared to the initial level (indoor contents: from 1.0 to 2.85; outdoor contents: from 1.0 to 3.9). Although this does not clearly demonstrate an overall improvement in cognitive function, it suggests that repetitive VR training can enhance task-specific performance in individuals with AD. Previous studies have shown that VR programs simulating real-life activities can improve abilities to carry out daily living activities in individuals with cognitive impairments.^49,50 Since the tasks in the Mentitree program are also based on everyday activities, this implies the need for further research on whether the program can support the performance of everyday activities in patients with AD.

This study has several limitations. First, the sample size was relatively small. With a larger sample size, the reliability of the study could be increased, and the influence of patient characteristics such as dementia severity or APOE genotype on the treatment could be identified through subgroup analyses. Second, the intervention period was relatively short (9 weeks). Third, no randomized control group was included. Therefore, future studies should include a larger sample size, administer long-term interventions, and set a randomized control group.

Conclusions

Cognitive training through VR intervention showed high feasibility and tolerability for patients with mild to moderate AD, suggesting that the program could be safely applied to real patients with minimal risks. It could be useful for the treatment of patients with AD by improving their visual recognition memory and helping maintain their cognitive function without showing substantial decline.