Files
Ovaert_83181300_Verstraete_25021500_2020.pdf
Open access - Adobe PDF
- 4.54 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- The Movement Disorder Society - Unified Parkinson’s Disease Rating Scale (MDS- UPDRS) is a scale used by physicians to assess the evolution of the Parkinson’s Disease (PD) in their patients. However, this assessment suffers from physician bias and variability. The biotech Tools4Patient, partner of this thesis, has developed the SensorMotor, a tool that records acceleration and angular velocity of specific hand movements of the MDS-UPDRS in order to predict more accurate and less variable scores to assess the progression of the disease. The first contribution of this thesis is to propose machine learning approaches to analyze these movement data. The second is to study the influence of the age and of the training effect on the recorded movements by using the proposed machine learning models to predict the age and the repetition number of a record. Needed steps to achieve these goals are to acquire several repetitions of good quality movement data through the development of a protocol, and to propose features extraction techniques to build an informative data set for the proposed machine learning models: Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP). Eventually, under the assumption that the training effect would be quite small, the following idea is proposed: summarize the data for the different repetitions of a person’s movement by keeping only the median value for each feature of the data set, in order to improve the age range predictions. Results show that the influence of the age and the training effect are quite low, but nevertheless present in the recorded movements. Moreover, the comparison of the performances of the different machine learning models shows that the use of the proposed summarized data allows to statistically improve the age range prediction of the best ranked classifier for movement data.