Abstract:
Wrist rigidity is accepted as a major indicator of the Parkinson’s Disease (PD). In the current medical practice, wrist rigidity is qualitatively measured by movement disorder specialists as a part of the Unified Parkinson’s Disease Rating Scale (UPDRS) method. This study aims to develop a novel method for quantitative assessment of wrist rigidity. In this thesis, a frequency analysis-based approach is proposed. Fre quency response of the wrist is obtained by analyzing data measured with the built-in sensors of a mobile phone attached to the hand. For this purpose, a mobile app and an experimental protocol are developed. An experimental study with 10 PD patients and 11 able-bodied participants is performed to correlate frequency response of the wrist with the UPDRS rigidity score. Along with the UPDRS rigidity score, differences be tween on and off states of PD patients, the effect of the Froment maneuver, differences between PD patients and healthy subjects are observed. Two analyses are performed with the collected data: (1) Model-based analysis, and (2) Deep learning-based anal ysis. A second-order mass-spring-damper model is fit to the data to identify stiffness of the wrist in the model-based analysis. On the other hand, a convolutional neural network is used to extract features related to wrist rigidity in the deep learning-based analysis. Statistical analysis of the results shows that it is possible to correctly identify UPDRS rigidity scores of the participants using the model-based analysis. Similarly, ac curate rigidity extraction is possible using the machine learning-based analysis. These promising results indicate that the proposed method along with the developed mobile app may be used for quantitative measurement of the muscle rigidity of PD patients.
