Development of Automated Gait Assessment and Intelligent Virtual Reality Based Gait Training Systems for Post-Stroke Rehabilitation

Stroke remains a leading cause of long-term disability, with approximately 80% of survivors experiencing gait impairments that severely compromise mobility, independence, and quality of life. This thesis aims to integrate advanced technologies, including artificial intelligence (AI) and virtual reality (VR) into post-stroke gait assessment and training, thereby developing solutions that are technically advanced, clinically useful, and implementable to support objective, efficient, and personalised rehabilitation. The overarching objective is to develop an automated gait assessment system and a VR-based gait training system for post-stroke patients, adhering to motor learning and control principles while promoting clinical use. Initial systematic and literature reviews identified challenges and gaps in system design and clinical application. Objective 1 developed and evaluated an AI-driven gait assessment system leveraging knowledge graph (KG) to analyse gait kinematic data, identify deviations, and dynamically pinpoint potential contributors (e.g., joint impairments). A preliminary evaluation using a twenty-post-stroke patient's gait dataset and four specialists demonstrated effective retrieval of relevant results and high acceptance. Objective 2 developed a low-cost VR-based treadmill training system for post-stroke patients, following a user-centred design framework that adhered to motor learning and control principles, with stakeholder considerations. Six exergames were developed with several adaptive controllers. Preliminary findings from physiotherapists and post-stroke patients showed good performance in adherence to motor learning principles, acceptability, and usability. To better understand the design of VR cues, this thesis further investigated how curved virtual paths in VR-based treadmill training influence gait symmetry and spatiotemporal parameters. A pilot study on older adults, the dominant stroke population, revealed that curved virtual paths significantly influence gait parameters including outer leg swing time, inner leg stance time, and centre of mass position. These findings provide preliminary evidence supporting future virtual cues design exploration in stroke rehabilitation. The work within this thesis underscores the potential of low-cost and efficient technology-driven solutions to optimise rehabilitation and empower patients, which is a critical step toward scalable, patient-centred healthcare. By aligning technical development with clinical relevance, this research lays a foundation for future exploration and gradual integration of intelligent rehabilitation systems into real-world practice.

Name thesis:degree_name

PhD

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Exercise Sciences

Grantor dc:publisher

ResearchSpace@Auckland

Year dc:date.issued

2025

Author dc:creator

• Jiao, Yiran

Advisors dc:contributor.advisor

• Zhang, Yanxin
• Reading, Stacey
• Smith, Marie-Claire

• Gait assessment
• Stroke rehabilitation

Statement dc:rights

• Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Licence dc:rights.uri

• https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm

Handle dc:identifier.uri

https://hdl.handle.net/2292/74298

OAI identifier oai:identifier

oai:researchspace.auckland.ac.nz:2292/74298