Wearable sensor technologies applied for post-stroke rehabilitation

Stroke is a common cerebrovascular disease that is recognized as one of the leading causes of death and ongoing disability around the globe. Stroke can lead to losses of various body functions depending on the affected area of the brain and leave significant impacts to the victim’s daily life. Post-stroke rehabilitation plays an important role in improving the life quality of stroke survivors. Properly designed rehabilitation training programs can not only prevent further functional deterioration, but also helps patients gradually regain their body functionalities. However, the delivery of rehabilitation service can be a complex and labour intensive task. In conventional rehabilitation systems, the chart-based ordinal scales are considered the dominant tools for impairment assessment and the administration of the scales primarily relies on the doctor’s manual observation. Measuring instruments such as strain gauge and force platforms can sometimes be used to collect quantitative evidence for some of the body functions such as grip strength and balance. However, the evaluation of the patients’ impairment level using ordinal scales still depend on the human interpretation of the data which can be both subjective and inefficient. The preferred scale and evaluation standard also vary among institutions across different regions which make the comparison of data difficult and sometimes unreliable. Furthermore, the intensive manual supervision and support required in rehabilitation training session limits the accessibility of the service as the regular visit to qualified hospital can be onerous for many patients and the associated cost can impose an enormous financial burden on both the government and the households. The situation can be even more challenging in developing countries due to higher growing rate of stroke population and more limited medical resources. The works presented in this thesis are focused on exploring the possibilities of integrating wearable sensor and pattern recognition techniques to improve the efficiency and the effectiveness of post-stroke rehabilitation by addressing the abovementioned issues. The study was initiated by a comprehensive literature review on the latest motion tracking technologies and non-visual based Inertia Measurement Unit (IMU) had been selected as the most suitable candidate for motion sensing in unsupervised training environment due to its low-cost and easy-to-operate characteristics. Following the design and construction of the 6-axis IMU based Body Area Network (BAN), a series of stroke patient motion data collection experiments had been conducted in conjunction with the Jiaxing 2nd Hospital Rehabilitation Centre in Zhejiang province, China. The collected motion samples were then investigated using various signal processing algorithms and pattern recognition techniques to achieve the three major objectives: automatic impairment level classification for reducing human effort involved in regular clinical assessment, single-index based limb mobility evaluation for providing objective evidence to support unified body function assessment standards, and training motion classification for enabling home or community based rehabilitation training with reduced supervision. At last, the study has been further expanded by incorporating surface Electromyography (sEMG) signal sampled during rehabilitation exercises as an alternative input to enhance accurate impairment level classification. The outcome of the investigations demonstrate that the wearable technology can play an important role within a tele-rehabilitation system by providing objective, accurate and often realtime indications of the recovery process as well as the assistance for training management