Evaluation of post-stroke hemiplegic patients is an important aspect of rehabilitation, especially for assessing improvement of a patient’s condition from a treatment. It is also a necessary step to perform during clinical trials. The Fugl-Meyer Assessment (FMA) is one of the most widely recognized and utilized measures of body function impairment for post-stroke patients. We propose a method for automating the upper-limb portion of the FMA by gathering data from sensors monitoring the patient. Features are extracted from the data and processed by a machine learning algorithm. The output from the machine learning algorithm returns a value that can be used to score a patient’s upper limb functionality. The machine learning algorithms tested in our system were Support Vector Machines (SVM) and Backpropagation Neural Networks (BNN). This system will enable automating and inexpensive stroke patient evaluation that can save up to 30 minutes per patient for a doctor, providing a time-saving service for doctors and stroke researchers. ⓒ 2015 DGIST
Table Of Contents
I.INTRODUCTION 1 -- II. RELATED WORKS 7 -- III. SENSOR DATA 10 -- 3.1 Kinect 11 -- 3.2 Wired Glov 12 -- 3.3 Pressure Sensor 12 -- 3.4 Additional Sensors 13 -- 3.5 Data Aggregation 15 -- IV. MACHINE LEARNING 16 -- 4.1 SVM Overview 16 -- 4.2 BNN Overview 18 -- 4.3 Feature Extraction 19 -- 4.4 Training SVM 24 -- 4.5 Training BNN 27 -- V. EXPERIMENTAL RESULTS 30 -- VI. CONCLUSION 32 -- References 33 -- Korean Summary 36 -- Acknowledgements 37 -- Curriculum Vitae 38