Estimation of Joint Angles and Moments of Rabbits from Electroneurograms using Muscle Spindle Model and Artificial Neural Networks
| Autor: | Ching-ChaoChan, 詹景超 |
|---|---|
| Rok vydání: | 2014 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 102 For subjects with spinal cord injuries, neural commands from the brain to the muscle system are blocked, and therefore there is a loss of voluntary action. Generally, treatment adopts conservative neural rehabilitation. An alternative is to use a prosthesis, which helps patients perform daily activities; however, the load and appearance of the prosthesis have led to low patient acceptability. Therefore, in recent years, researches proposed using implantable functional electrical stimulation system in which electroneurograms (ENGs) has been used to extract information of body motion for use as a feedback signal in controlling current and stimulating muscle contractions to restore motor function. The system could replace external components completely. Currently, ENG feedback methods only used skin sensory signals to extract tactile and slip information in on-off control. Body motion information is not applied for feedback control. The aim of this study is to estimate ankle joint angle and moment by extracting proprioception signals which are related to muscle length and force. In this study, it is focused on developing functional electrical stimulation system of rabbit ankle joint. A method for estimating ankle joint angles from ENGs without muscular stimulation is first developed. Muscle spindle models with physiological structure and large-range joint movements were used to construct two models that relate joint angle to tibial and peroneal ENGs. Then, the two estimated angles were calculated by a combiner to obtain a final angle estimate. Besides, a neuro-fuzzy system is also developed for accelerating parameter learning process. When active joint moment is estimated, passive joint moment is needed which comes from all tissues surrounding the joint. The existing passive moment models cannot describe the steady-state hysteretic behavior of joint angle and moment. Therefore, the Preisach-based model was adopted to build the relationship between passive joint moment and angle. For muscle contraction, the agonist-antagonist muscles were activated by electrical stimulation to obtain the active joint moments at different joint angles and then two Hill-type models were built to relate stimulus intensity and joint angle. When the muscles are contracted by electrical stimulation, their ENGs may contain Golgi tendon organ signals or interference from the electrical stimulation. Therefore, the ENG was hypothesized to be the algebraic sum of muscle spindle and Golgi tendon organ components. Then the active-moment-related ENG models were then constructed. Finally, a system architecture with physiological basis to relate ankle joint moment, angle and ENGs was constructed, in accordance with the above-mentioned models and the hypothesis of ENG separability. Then, an ankle motion estimator capable of simultaneously estimating ankle joint angle and moment under electrical stimulation was built. In vivo experiments of 12 New Zealand rabbits were performed to validate the above-mentioned system architecture. The experimental results showed that the muscle spindle method outperforms the neuro-fuzzy system in terms of angle estimation and it can be used on large-range movement. It was indirectly verified that ENG changes recorded during passive stretches mainly come from the muscle spindle response. The combiner of the model which was used to predict joint movement direction could eliminate the hysteretic error between ENGs and angle. Besides, the neuro-fuzzy system had a higher parameter learning speed, and was thus considered satisfactory for online applications. For the passive moment, the Preisach-based model can reduce estimation error caused by the moment-angle hysteresis. After integrating with the active moment model, better total moment can be estimated. For isometric muscle contraction, the active-moment-related ENG can be used to estimate the active moment. Finally, the ankle motion estimator can simultaneously estimate ankle joint angle and moment when subjected to electrical stimulation. This study proposed the hypothesis of ENG separability and constructed the relationship among ankle joint angle, moment and ENGs based on physiology of muscle spindle and dynamics of muscle subjected to electrical stimulation. By integrating all the models, the estimation of ankle joint angle and moment was achieved successfully. In the future, the method might be used in realizing completely implanted functional electrical stimulation system. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|