Augmented reality-assisted Tai Chi for adults with diabetes-associated cognitive impairment: study protocol for a randomised controlled trial

Abstract

Introduction

Diabetes-associated cognitive impairment (DACI) undermines diabetes self-management and is associated with increased morbidity and mortality. Although mind-body exercise may improve cognition, its benefits are often limited by the lack of timely, precise, individualised instruction and feedback in adults with cognitive impairment. Augmented reality (AR) systems can deliver performance-adaptive, visually anchored cues and auditory prompts in real-world settings, but their cognitive and metabolic benefits in DACI remain unknown. This trial will compare AR-assisted Tai Chi (iTaiChi), instructor-led Tai Chi (TaiChi) and standard care in adults with DACI and explore underlying metabolic and neurobiological mechanisms.

Methods and analysis

In this randomised controlled trial, 246 community-dwelling adults aged 45–75 years with DACI will be enrolled and randomly allocated in a 1:1:1 ratio to a TaiChi group, an iTaiChi group or a standard care group. All participants will receive standard care. In addition, participants assigned to the two exercise groups will undertake a standardised Yang-style 24-form Tai Chi programme three times weekly for 12 weeks. The primary outcome is the change in Montreal Cognitive Assessment Score. Secondary outcomes include physical health related parameters and brain health related parameters.

Ethics and dissemination

This study was approved by the Ethics Committee of Shanghai Sixth People’s Hospital (Approval No. 2025–017). Dissemination of the findings will include peer-reviewed publications, conference presentations and media releases.

Trial registration number

ChiCTR2500111614.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This three-arm randomised controlled trial in adults with diabetes-associated cognitive impairment will evaluate the incremental effect of augmented reality (AR)-assisted Tai Chi.

• A purpose-built AR system will provide real-time visual and auditory cues to deliver standardised, spatially anchored movement guidance.

• Multidimensional data will be collected, including 3T MRI and multiomics, to explore mechanisms underlying intervention effects.

• A limitation is that participants cannot be blinded because of the nature of the exercise interventions, although outcome assessors will remain blinded.

Introduction

Type 2 diabetes is a major global public health challenge, with prevalence continuing to rise worldwide.¹ An increasingly recognised neurocognitive complication of type 2 diabetes is diabetes-associated cognitive impairment (DACI). The prevalence of mild cognitive impairment (MCI) among individuals with type 2 diabetes has been reported to be as high as 45%.² Moreover, among those already diagnosed with MCI, type 2 diabetes has been linked to faster conversion to dementia, with progression rates reported to be approximately 1.5–3.0 times higher than in people without type 2 diabetes.³ Cognitive impairment compromises diabetes self-management capacity, reduces treatment adherence and worsens quality of life, contributing to increased morbidity and mortality.⁴

Mind-body exercises such as Tai Chi integrate physical movement with attentional engagement and coordinated motor control and have been associated with improvements in metabolic health and cognitive performance in older adults,^5–7 with neuroimaging studies suggesting potential influences on brain function and connectivity.^{8 9} Meta-analyses report that such interventions improve glycaemic control, lipid metabolism and insulin resistance in individuals with type 2 diabetes while also enhancing cognitive functions, including executive function, attention and memory.^{7 10} These benefits depend on consistent training exposure and accurate movement execution.^{11 12} However, in real-world practice, mind-body exercise programmes are typically delivered through instructor demonstration and verbal cueing, limiting continuous, individualised feedback and introducing variability in training exposure.^{13 14} These limitations are particularly relevant to adults with DACI.

Augmented reality (AR)-based technologies may offer a potential approach to addressing these delivery limitations by providing real-time visual and auditory guidance during exercise practice.^{15 16} Previous research on immersive digital exercise interventions has also suggested potential benefits for cognitive outcomes.¹⁷ In adults with DACI, impairments in attention, memory and executive function may reduce the effectiveness of verbal-only exercise instruction,¹⁸ whereas multimodal cueing and external prompting may facilitate task understanding and support cognitive-motor performance.¹⁹ Evidence from rehabilitation research suggests that AR-assisted exercise systems can improve motor performance and functional outcomes,²⁰ and clinical studies have also reported good feasibility and adherence in AR-assisted home rehabilitation programmes.²¹ In mind-body exercise contexts, a randomised trial reported that AR-assisted practice of selected movements improved balance control and lower-limb strength in community-dwelling older adults.²²

However, the effects of AR-assisted exercise on cognitive function and metabolic health in adults with DACI remain unknown. It is unclear whether AR-assisted exercise confers an incremental benefit over instructor-led exercise in this population. In addition, mechanistic evidence linking intervention exposure to neurobiological and metabolic changes in adults with DACI remains limited.

Objectives

In this study, Tai Chi was selected as the mind-body exercise paradigm because it is a representative and well-studied modality in adults and has been reported to confer benefits for both cognitive and physical function in ageing populations.^{5 23} This study aims to evaluate the effects of instructor-led Tai Chi (TaiChi) and AR-assisted Tai Chi (iTaiChi), compared with standard care, in adults with DACI and to explore potential mechanisms through which exercise interventions, enhanced by AR-assisted spatial and multimodal guidance, may influence cognitive, metabolic and brain-related outcomes. These assessments will include: (1) Cognitive function related measures; (2) Physical health related measures; (3) Brain health related parameters; and (4) Multiomics analyses.

Methods and analysis

Study design

This is a randomised controlled trial in adults with DACI. The study protocol is reported in accordance with the Standard Protocol Items: Recommendations for Interventional Trials 2025 statement.²⁴ Participants will be randomly allocated to three groups: (1) TaiChi group, (2) iTaiChi group and (3) Standard care group.

All participants will undergo a 12-week intervention and attend two study visits, one at baseline and the other at the end of the intervention period, as shown in figure 1. At each visit, blood, saliva, urine and stool samples will be collected, and physical and brain health related parameters will be assessed. Postintervention samples will be collected approximately 48 hours after the final exercise session. Participants will also be instructed to avoid vigorous physical activity during this interval before sample collection. The study was registered with ChiCTR in November 2025 (ChiCTR2500111614). It began on 15 November 2025 and is expected to be complete on 30 May 2026.

Figure 1. Overview of the study design. After screening, eligible participants will be randomly assigned to three groups for a 12-week intervention. The TaiChi group will receive supervised exercise sessions delivered by certified instructors three times per week, while the iTaiChi group will perform the same exercise programme using AR glasses equipped with the iTaiChi system at the same frequency. The standard care group will receive standard diabetes care. Outcome assessments will be conducted after 12 weeks. AR, augmented reality; CGM, continuous glucose monitoring; iTaiChi, AR-assisted Tai Chi; TaiChi, instructor-led Tai Chi; TCM, Traditional Chinese Medicine.

Participants and eligibility criteria

A total of 246 adults with DACI will be recruited from community settings in the Nicheng area of Shanghai. Type 2 diabetes in this population will be diagnosed according to the diagnostic criteria of the WHO, and MCI will be diagnosed based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition.²⁵ Detailed eligibility criteria are listed in table 1.

Table 1. Eligibility criteria.

Inclusion and exclusion criteria
Inclusion criteria
1	Diagnosed with DACI
2	Aged 45–75 years
3	Able to understand and complete the MoCA and other relevant assessments
4	On a stable glucose-lowering regimen for at least 2 months before enrolment
5	Written informed consent provided by both the participant and their legally authorised representative
Exclusion criteria
1	Diagnosis of other types of diabetes, including type 1 diabetes or gestational diabetes
2	Presence of major neurological disorders, including Alzheimer’s disease, Parkinson’s disease, stroke sequelae, traumatic brain injury, intracranial haemorrhage or acute cerebral infarction
3	Severe cardiopulmonary disease, such as uncontrolled hypertension or coronary artery disease
4	Use of medications that may impair cognitive function, including certain anticonvulsants and antipsychotics
5	Dementia, major psychiatric disorders or a history of major depression within the past 2 years
6	Severe sinusitis, nasal or sinus polyps, space-occupying lesions such as skull base or nasopharyngeal tumours, or congenital or traumatic disorders affecting the olfactory system
7	Acute diabetic complications, including diabetic ketoacidosis, hyperosmolar hyperglycaemic state or hypoglycaemic coma
8	Contraindications to MRI, including metallic implants or claustrophobia
9	Active upper respiratory infection on the day of MRI, including symptoms such as nasal congestion, rhinorrhoea or fever
10	Pregnancy or breastfeeding
11	Severe dysfunction of major organs, including the heart, liver or kidneys
12	Significant visual or auditory impairment that interferes with communication or participation in exercise training
13	Muscle, bone, nerve or other tissue and organ damage that may hinder exercise training
14	Participation in other clinical trials that may potentially affect the outcomes of this study

DACI, diabetes-associated cognitive impairment; MoCA, Montreal Cognitive Assessment.

Interventions

All eligible participants will be randomly assigned to the TaiChi group, iTaiChi group or standard care group. All groups will receive standard care, such as guidance on a healthy diet, physical activity, blood glucose monitoring and glucose-lowering medication.^{26 27}

Exercise intervention

Both exercise groups will receive supervised training three times per week for 12 weeks, with each session lasting 60 min. All sessions will be conducted by experienced Tai Chi instructors trained in the study protocol. Training adherence, session completion, and any injuries or adverse events (AEs) will be recorded. Caregivers or guardians may accompany participants to the sessions if needed.

Trained medical personnel will be present during all supervised exercise sessions. Before each session, the exercise area will be checked, participants’ general condition will be assessed and capillary blood glucose will be measured to ensure safe exercise participation. In accordance with clinical guidance,^{26 28 29} explicit pre-exercise blood glucose safety thresholds and corresponding actions will be applied. If blood glucose is <3.9 mmol/L (70 mg/dL), exercise will be withheld, fast-acting oral carbohydrate will be provided and blood glucose will subsequently be rechecked, with further treatment given as necessary. If blood glucose is >16.7 mmol/L (300 mg/dL), the session will be postponed and the participant will be assessed for symptoms of hyperglycaemia, dehydration or ketosis. Any signs of acute metabolic instability will preclude exercise and prompt medical evaluation.

Throughout each session, trained medical personnel will continue to monitor participants’ general condition and exercise tolerance, while experienced Tai Chi instructors will supervise movement performance and designated technical staff will oversee system operation. For participants in the iTaiChi group, discomfort related to AR device use, including dizziness, nausea, headache, blurred vision, eye strain, visual discomfort, disorientation, motion sickness-like symptoms, imbalance or excessive fatigue, will also be assessed. If any of these signs or symptoms occur, the session will be interrupted immediately. The participant will then be assessed by trained medical personnel, who will determine whether the session can be safely resumed or should be terminated. All such events will be documented.

After each exercise session, participants will undergo a 15-min observation period to ensure safe completion of the session and will be discharged only if no abnormalities are identified. Any immediate or delayed discomfort will be managed according to the study safety procedures and recorded.

TaiChi group

Participants in the TaiChi group will receive the Yang-style 24-form Tai Chi programme delivered on-site by professional Tai Chi instructors. Each practice session will include a 10-min warm-up, 40-min Tai Chi practice and a 10-min cool-down. Exercise training will be progressive.

iTaiChi group

Participants allocated to the iTaiChi group will wear AR glasses equipped with the iTaiChi system and follow a virtual instructor presented through the AR interface. Before starting the intervention, participants will complete a brief familiarisation session with the AR display and exercise interface. The iTaiChi system was developed by our team to deliver a standardised Tai Chi programme with real-time instructional guidance. It is designed to support exercise practice in adults by providing multimodal feedback while minimising the need for additional wearable sensors or continuous instructor-led correction.

The intervention in the iTaiChi group will be delivered using commercially available Rokid Max 2 AR glasses. According to the product manual, the device weighs 75 g and uses dual Sony Micro-OLED display panels, with a monocular resolution of 1920×1080 pixels, binocular full high-definition display, a 50° field of view and a peak perceived brightness of 600 nits. The iTaiChi software architecture is designed to be adaptable to other commercially available AR glasses that meet the required display and data-transmission specifications.

During exercise, participants will wear the AR glasses and train within a predefined physical exercise area. An external monocular RGB camera will be placed at a fixed position outside the training area to capture whole-body movements without obstructing exercise practice. The captured movements will be processed on an external computer, which will generate feedback and transmit it to the AR glasses through compatible audiovisual interfaces. Through the AR interface, participants will follow a virtual Tai Chi instructor and receive multimodal guidance, including visual prompts for movement direction, target position, movement trajectory and safe training boundaries, together with auditory prompts for movement instructions, pacing, verbal encouragement and corrective feedback. Real-time corrective feedback will then be delivered through visual cues in the AR interface and synchronised auditory prompts (figure 2).

Figure 2. Conceptual framework of the iTaiChi system. Participant movements are captured in real time using a camera-based system. Body keypoints are extracted to identify deviations from reference Tai Chi movements demonstrated by a virtual instructor. Based on the detected deviations, the system provides real-time multimodal feedback, including spatial guidance and visual and auditory cues, to guide posture correction during exercise practice. iTaiChi, AR-assisted Tai Chi.

Under a fixed camera configuration and within a predefined training area, the monocular RGB pose-estimation pipeline estimates selected body landmarks and derives body-centred spatial features for phase-specific assessment of movement direction, foot placement, gross postural stability and movement sequence. To mitigate transient tracking degradation caused by self-occlusion and profile views, landmark trajectories are temporally filtered and subjected to confidence-based reliability checks. Automated movement assessment is performed only when the landmarks required for the corresponding movement phase meet the predefined reliability criteria; otherwise, deviation classification and corrective feedback are withheld for that phase. This reliability-gated, phase-level assessment reduces dependence on continuous frame-by-frame joint-angle matching and limits false-positive feedback during normal movement transitions and brief occlusions.

In iTaiChi, standard Tai Chi movements were captured as a continuous skeleton trajectory and segmented into short movement phases. During training, the participant’s body-centred and body-scale-normalised skeleton sequence was aligned in real time with the corresponding reference phase. The normalised Euclidean landmark distance was used to assess frame-level trajectory concordance.^{30 31} In the instructional mode, each phase was practised separately with a virtual-instructor demonstration, audio explanation and visual guidance. Performance was assessed throughout using prespecified criteria for trajectory concordance and movement execution, with endpoint attainment assessed at phase completion. Meeting the applicable criteria triggered immediate encouragement; otherwise, an appropriate corrective cue and a repeat demonstration were provided before progression. In the practice mode, phases were linked into a continuous sequence performed at a slower pace and assessed against mode-specific criteria. Meeting the criteria triggered an encouraging audio message; otherwise, brief visual and auditory cues were delivered without interrupting the sequence, followed by concise verbal feedback.

To assess the practicality, functional visual safety and AR tolerability of this approach in individuals with DACI, a preliminary feasibility assessment was conducted in 20 participants with DACI. The assessment was also used to further refine the movement-specific tolerance ranges and task windows based on participants’ performance and completion times during representative Tai Chi tasks. During the assessment, trained medical personnel remained beside each participant to provide individual supervision and ensure immediate safety. The assessment focused on three aspects: (1) Functional visual awareness, defined as the ability to recognise safety boundaries and visible obstacles and to navigate safely within the predefined training area while wearing the AR glasses; (2) AR tolerability, defined as the absence of dizziness, blurred vision, eye strain, visual discomfort, disorientation, nausea or imbalance; and (3) Instruction-following ability, defined as the ability to safely follow visual or auditory prompts and complete basic Tai Chi movements within the allotted time.

Participants were stratified by cognitive status using the Montreal Cognitive Assessment (MoCA) Score. In this assessment, 11 of 12 participants (91.7%) with MoCA scores ≥18 and 6 of 8 participants (75.0%) with MoCA scores <18 were able to independently complete the core AR-assisted training procedure. The remaining three participants met the functional visual awareness and AR tolerability criteria but did not complete all required movement tasks within the allotted time and were therefore classified as not having completed the procedure independently. No serious adverse events (SAEs) were reported during system use. These preliminary results supported the feasibility, functional visual safety and AR tolerability of implementing the AR-assisted intervention in individuals with DACI.

Process evaluation

Process evaluation will be used to collect data on the implementation of the intervention, contextual factors and participants’ motivation. We will interview any participant who chooses to discontinue their involvement in the study to investigate the factors contributing to their decision to withdraw. In addition, the protocol outlines procedures to support participant retention, ensure complete follow-up and specify the outcome data to be collected from participants who discontinue or deviate from the intervention.

Outcomes

Primary outcome

The primary outcome will be the change in MoCA Score.

Secondary outcomes

Physical health related outcomes

Physical health related outcomes will include the following: (1) Anthropometric indicators such as body weight, body mass index, waist circumference, hip circumference, thigh circumference, neck circumference, etc; (2) Body composition, including fat mass, fat-free mass, muscle mass and body fat percentage, etc; (3) Metabolic and clinical biochemical indicators, including glucose and lipid metabolism indicators, kidney function and liver function, etc; (4) Sarcopenia-related indicators, including the Strength, Assistance with walking, Rise from a chair, Climb stairs, and Falls questionnaire combined with calf circumference, calf circumference measurement, the five-times sit-to-stand test, 6-metre walk test and handgrip strength test, etc; (5) Exercise-related indicators, including muscle strength test, muscle endurance test, cardiorespiratory endurance test, Global Physical Activity Questionnaire, etc; (6) Continuous glucose monitoring derived metrics; (7) Psychological status and sleep quality, such as Patient Health Questionnaire-9, Self-Rating Anxiety Scale, University of California at Los Angeles Loneliness Scale and Pittsburgh Sleep Quality Index, etc. All questionnaires will be administered by trained physicians using a standardised interview-based approach, with responses recorded based on participants’ own answers.

Brain health related outcomes

Brain health related parameters will include the following: (1) Brain MRI scans will be performed on a 3T United Imaging UMR890 scanner with a 64-channel head coil, and the imaging protocol will comprise three sequences: structural MRI (sMRI), resting-state functional MRI (fMRI) and task-based fMRI; (2) Cognitive subdomain tests will include the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), the Digit Span Test (forward and backward), the Trail Making Test and the Stroop Color-Word Test; (3) Executive function assessment, including the N-back task (for working memory assessment) and the Flanker task (for inhibitory control assessment), will be conducted; (4) Olfactory function will be assessed using the olfactory identification test.

Exploratory outcomes

Multiomics analysis

Multiomics analysis will include metagenomics, metabolomics, genomics, lipidomics, proteomics, etc.

Traditional Chinese Medicine related features

Traditional Chinese Medicine related features will be recorded using a standardised instrument-based assessment system. The outcome measures and assessment protocols are listed in table 2.

Table 2. Outcome measures and assessment protocols.

Measurements	Time points		Method of assessment
	Baseline	12 weeks
Primary outcome
MoCA Score	•	•	MoCA questionnaire
Secondary outcomes
Physical health related parameters
Anthropometric indicators	•	•	Anthropometric measurement
Body composition	•	•	Bioimpedance analysis
Metabolic and clinical biochemical indicators	•	•	Biomedical test
Sarcopenia-related indicators	•	•	SARC-CalF Scale, calf circumference measurement, five-times sit-to-stand test, 6-metre walk test and handgrip strength test
Exercise-related indicators	•	•	Muscle strength test, muscle endurance test, cardiorespiratory endurance test, GPAQ
CGM-derived metrics	•	•	CGM data
Psychological status and sleep quality	•	•	Psychological and sleep questionnaires
Brain health related parameters
Brain MRI	•	•	sMRI, rs-fMRI, task-based fMRI
Cognitive subdomain tests	•	•	RBANS, Digit Span Test, TMT, Stroop Color-Word Test
Executive function	•	•	N-back and Flanker tasks
Olfactory function	•	•	Olfactory identification test
Exploratory outcomes
Multiomics analysis
Metagenomics	•	•	Metagenomic sequencing
Metabolomics	•	•	LC-MS/MS
Genomics	•	•	GWAS
Lipidomics	•	•	LC-MS/MS
Proteomics	•	•	LC-MS/MS
TCM-related features
TCM-related measures	•	•	Standardised instrument

CGM, continuous glucose monitoring; fMRI, functional MRI; GPAQ, Global Physical Activity Questionnaire; GWAS, Genome-wide Association Study; LC-MS/MS, liquid chromatography-tandem mass spectrometry; MoCA, Montreal Cognitive Assessment; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; rs-fMRI, resting-state functional MRI; SARC-CalF, Strength, Assistance with walking, Rise from a chair, Climb stairs, and Falls Questionnaire combined with calf circumference; sMRI, structural MRI; TCM, Traditional Chinese Medicine; TMT, trail-making test.

Adverse events

AEs will be monitored throughout the trial. Anticipated AEs include musculoskeletal symptoms or injuries, falls, symptoms suggestive of glycaemic fluctuation such as suspected hypoglycaemia, and AR device related discomfort such as dizziness, nausea, headache, visual discomfort, eye strain, motion sickness-like symptoms or skin irritation. AEs will be identified through participant self-report, direct questioning and observation by medical personnel and exercise instructors, and will be recorded in case report forms (CRFs) and the study database, including onset, resolution, severity and relatedness to the intervention.

AEs will be classified as mild, moderate or severe. Mild AEs will be managed by the study team, moderate AEs may require clinical assessment and temporary interruption of the intervention, and severe AEs will be promptly evaluated by the study clinician, with further participation determined according to clinical judgement and participant safety. SAEs will include events resulting in death, life-threatening conditions, hospitalisation or prolongation of hospitalisation, or persistent or significant disability, and will be reported to the ethics committee within 24 hours. Resumption of the intervention after any AE will be determined by the study clinician following safety assessment.

Sample size

Sample size was calculated for the primary outcome (MoCA Score change from baseline to 12 weeks). Based on our preliminary data, we assumed a between-group difference of 1.5 points in MoCA Score with an SD of 2.7. Because this is a three-arm trial with three prespecified pairwise primary comparisons, a two-sided significance level of α=0.05/3 (0.0167) and 80% power were used. Under these assumptions, the minimum required sample size is 69 participants per group. Considering a 15% attrition rate, we will recruit 82 participants per group, for a total sample size of 246 participants.

Recruitment

Residents aged between 45 years and 75 years in the Nicheng Area of Shanghai will be recruited through flyers, promotional videos and health lectures. Potential participants will be identified by medical experts, who will act as recruiters. Physicians will assess the residents’ blood glucose levels and cognitive function to identify individuals with DACI and verify that they meet the aforementioned eligibility criteria. All participants will be informed of the study’s objectives and key features. Informed consent will be obtained from all participants and their legally authorised representative (see online supplemental file).

Allocation, concealment and blinding

Participant group allocation will be determined by random numbers generated using SAS PROC PLAN in SAS software, with numbers stratified by sex and age group. On obtaining participants’ consent, they will be assigned to the three groups at a 1:1:1 ratio. Given the nature of this exercise intervention study, blinding of participants to treatment cannot be implemented. However, all assessments will be conducted by blinded independent assessors, and participants will be instructed not to disclose their group allocation during testing visits. Participants’ unique study identification codes will serve as the sole identifier for data storage, and any data analyses susceptible to researcher bias will be performed without the knowledge of intervention allocation.

Data management and analysis

Data management

A CRF will be used in this trial to collect information and will be kept strictly confidential for research purposes. The CRF for each participant will be filled out by trained researchers who are blinded to the group assignments. Researchers will keep files intact, store them in a fixed place, and lock them at the Department of Endocrinology and Metabolism, Shanghai Sixth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine. All data will be scrutinised and double-entered by the researchers. Before database locking, all demographic and clinical variables will be reviewed against predefined variable definitions and data dictionaries during the audit process. Behavioural symptoms, questionnaire-derived measures, adverse symptoms and established medical disorders will be classified into distinct variable categories in the primary database. After the database is established and confirmed through a blind audit, it will be locked, and no further change will be permitted.

Statistical methods

Data analysis for this study will be conducted using IBM SPSS Statistics V.27.0. Statistical analysis will follow the intention-to-treat principle, including all randomised participants in the groups to which they were originally allocated. Missing data will be handled using multiple imputation. Data will be presented as means±SD, median (IQR) and frequencies (percentages). The Kolmogorov-Smirnov test will be used to assess the normality of the data distribution.

For baseline characteristics, comparisons will be made using analysis of variance for normally distributed data, the Kruskal-Wallis H test for non-normally distributed data, and the χ² test for categorical data. Paired t-tests and Wilcoxon signed-rank tests will be used to compare baseline and follow-up values for normally and non-normally distributed data, respectively. Changes in clinical and multiomics data before and after the intervention will be compared between groups using covariance analysis, with baseline variables as covariates. The comparison will be conducted between the TaiChi group and the standard care group, the iTaiChi group and the standard care group, as well as the TaiChi group and the iTaiChi group. A Bonferroni correction will be applied for multiple testing of the primary outcome. With the exception of the primary outcome measure, whose significance will be ascertained using an adjusted two-tailed p<0.0167, all remaining statistical tests will be executed with a two-tailed significance level fixed at p<0.05, unless otherwise specified. Multiomics data, such as metagenomic data, will be analysed using MetaPhlAn software. The Wilcoxon test will be used to compare microbiota abundance before and after the intervention, while differences in microbiota abundance between groups will be assessed using the Mann-Whitney U test. Spearman’s correlation and partial correlation analyses will examine the relationships between microbiota abundance, metabolite levels and clinical indicators.

Monitoring

An independent Data Safety and Monitoring Board (DSMB), composed of independent experts, will ensure data management reproducibility and analysis integrity. The DSMB will review data collection, participant safety, protocol compliance and trial oversight, while also validating key processes, such as enrolment, consent, randomisation, intervention adherence and follow-up.

Patient and public involvement

Participant and family feedback will be sought and considered for minor protocol adjustments throughout the clinical trial. During the development phase, professional coaches contributed to the design of AR exercises. Following the creation of the initial AR exercise version, iterative refinements were made based on feedback to optimise functionality and user experience. This iterative process aimed to enhance immersion while minimising motion-induced discomfort associated with AR environments.

Ethics and dissemination

This study was approved by the Ethics Committee of Shanghai Sixth People’s Hospital (Approval No: 2025–017) and registered in the Chinese Clinical Trial Registry (ChiCTR2500111614). Written informed consent will be obtained from all participants and their legally authorised representatives. Any modifications made to the study protocol will be shared with the Ethics Committee of Shanghai Sixth People’s Hospital as stipulated, and we will follow their advice regarding implementation and dissemination. Dissemination of the findings will include peer-reviewed publications, conference presentations and media releases.

Discussion

This study is a three-arm parallel randomised controlled trial designed to compare the effects of iTaiChi, TaiChi and standard care in adults with DACI. Beyond cognitive function, assessed using the MoCA, the study aims to investigate whether AR-assisted exercise may confer multidimensional benefits spanning physical fitness, metabolic health and brain-related measures that collectively support cognitive improvement. By integrating these outcomes, this trial may help clarify the mechanisms through which AR-assisted exercise contributes to broader functional and systemic health benefits in adults with DACI.

The iTaiChi system extends beyond simple digitalisation of exercise delivery by integrating AR-assisted spatial guidance, multimodal visual and auditory cues, real-time motion feedback and an adaptive virtual instructor to facilitate exercise execution and monitoring. Such a design may be particularly relevant for individuals with DACI, for whom impairments in processing verbal instructions and maintaining accurate movement patterns may reduce the effectiveness of conventional exercise instruction. In addition, this trial employs a three-group comparative design including standard care, Tai Chi and iTaiChi, enabling evaluation not only of the effects of Tai Chi itself but also of the potential added value of AR-enhanced delivery beyond conventional instructor-led training.

For individuals with DACI, the cognitive assessment employed in this study integrates the pathological characteristics of diabetes with cognitive function domains, adopting a multimethod comprehensive approach to fully reflect the cognitive status. Routine screening measures will include the MoCA. We will also use RBANS, the Digit Span Test, and the Stroop Color-Word Test to conduct a comprehensive quantitative evaluation of multidimensional cognitive functions. Additionally, olfactory function maintains close anatomical and functional connections with core cognitive brain regions (entorhinal cortex, hippocampus, prefrontal cortex), and olfactory impairment often emerges prior to the onset of typical cognitive symptoms.^{32 33} Thus, olfactory assessment is incorporated to evaluate adults with cognitive impairment in this study. Furthermore, the N-back task and Flanker task will be applied to perform a refined assessment of specific cognitive functions, including working memory, inhibitory control and cognitive conflict resolution.^{34 35}

With the rapid development of brain MRI and multiomics technologies, new opportunities have emerged to investigate the mechanisms underlying DACI.^36–38 sMRI can provide measures of brain morphology, such as regional volume and cortical thickness, whereas fMRI can assess patterns of brain activity and functional connectivity that may be relevant to cognitive outcomes. Previous studies have reported that patients with DACI may exhibit structural and functional abnormalities, including hippocampal and frontal atrophy, white matter hyperintensities, microvascular lesions and altered functional connectivity.^{39 40} Emerging evidence also suggests that gut microbiota related pathways and systemic metabolic disturbances, particularly in amino acid, energy and lipid metabolism, may contribute to diabetes-related cognitive decline.^41–43 In this study, MRI and multiomics approaches will be used to explore associations between intervention exposure and neurobiological or metabolic measures.

Several limitations should be acknowledged. Head-mounted AR devices may cause transient discomfort in some participants,^{44 45} and the operation of the system requires a short learning period that may affect early adherence. The 12-week intervention is expected to capture short-term cognitive and biological changes, but longer follow-up is needed to understand durability and downstream clinical events. Moreover, participants with limited digital literacy or visual impairments may face usability challenges, potentially affecting adherence and generalisability. Finally, the neurocognitive mechanisms underlying iTaiChi remain to be elucidated in future multimodal and longitudinal studies.

In conclusion, this protocol outlines an innovative approach to promoting cognitive engagement in adults with DACI through AR-assisted exercise. By providing structured guidance and spatially immersive multisensory cues, AR technology may improve engagement with exercise practice and support task-focused participation during training. This trial will evaluate iTaiChi delivery compared with conventional delivery and standard care, while incorporating exploratory neuroimaging and multiomics measures to characterise associated biological profiles. This study is expected to provide exploratory evidence that may inform the design of future technology-assisted lifestyle interventions and may contribute to an evidence-based framework for integrating AR-assisted exercise into comprehensive management strategies targeting both cognitive health and metabolic health in this vulnerable population.