| Parkinson’s disease (PD) is a common neurodegenerative disease in middle-aged and elderly people. Tremor, muscle rigidity, bradykinesia and postural balance disorder are its four main symptoms. These symptoms reduce the patients’ quality of daily life in various degree, and severe patients even lose motor ability and can not live on their own. Moreover, the patients in advanced stages may appear dyskinesia and ON/OFF fluctuations of motor performance with the long-term levodopa treatment. Complexity and variability of the motor symptoms of PD make clinical assessment become a difficult task. The assessment tools of movement disorders widely used in clinical practice are the various assessment scales, which rely highly on raters’ clinical experience. It’s hard to ensure the accuracy and objectivity. Therefore, an effective objective assessment method is needed to determine the severity of motor impairment in order to help doctors to make accurate diagnois and evaluate the efficacy.The purpose of this thesis is to study the quantitative parameters and grading evaluation methods of typical upper limb motor symptoms for PD patients based on the motion signals of inertial sensors. The main work of this thesis is as follows:(1) The application and research status of inertial sensors in evaluating movement disorders of PD were deeply investigated. On the basis of early clinical tests with clinicians, a quantitative assessment system for upper limb movement disorders was established.(2) Design a motion signal acquisition system to collect 3D acceleration and 3D angular velocity by using inertial sensors. A database of different movement signals of PD patients was built to quantitatively analyze and evaluate the motor symptoms.(3) Filtering tremor component from the finger movement signals. A filtering algorithm based on Hilbert-Huang Transform was proposed to remove the tremor composition from finger taps movement signals in PD patients. First, the inertial sensor signals were decomposed by EMD, and then modal selection algorithm based on prior knowledge was designed. The selected modals were used to reconstruct the target movement signals. The effect of this method was compared with the traditional linear low-pass filter in practical clinical test signals.(4) Quantitative assessment of arm tremor symptoms. For resting tremor and postural tremor, the amplitude and frequency related characteristic parameters on time domain and frequency domain were extracted. The correlation between them and the clinical UPDRS scores of tremor items was analyzed. Classification models were used to grade the severity of two kinds of tremor. The outcomes of this algorithm were compared before and after the dimension reduction of feature space using PCA.(5) Quantitative assessment of upper extremity symptoms of bradykinesia. For finger repetitive movements, motor signals were collected when PD patients repetitively performed finger tapping and grasping movements. Classic time domain and frequency domain features were extracted. Their correlation with clinical UPDRS corresponding item scores was analyzed to validate their quantitative effectiveness of bradykinesia severity. Two nonlinear indexes of approximate entropy and sample entropy were introducted to quantify the rhythmicity of finger tapping movement.Finally, all above features were used to rate scores for two actions by a classification model. For both hands alternating movement, the patients’ bilateral wrist motor signals were collected. The characteristic parameters of left and right hand were calculated respectively and fused together for training a classification model to output rating score.(6) Comprehensive assessment of upper limb movement disorders. On the basis of previous research, test the predicting performance of classification models for each item on test set. All individual scores were summed to get model estimated total score for upper limbs. Then model estimated total score for ensemble movement disorders was returned by linear regression. |
PDF Full Text Request