Parameter-optimized non-invasive speech test for Parkinson's disease Severity Assessment
Parkinson's disease (PD) has a significant impact on people's lives following Alzheimer's, the most common neurological condition. It causes some or all loss of the ability to move, speak, behave, think, and perform other essential tasks. Monitoring the patient's condition on a f...
Saved in:
| Published in: | 2023 Eighth International Conference on Science Technology Engineering and Mathematics (ICONSTEM) pp. 1 - 7 |
|---|---|
| Main Authors: | , |
| Format: | Conference Proceeding |
| Language: | English |
| Published: |
IEEE
06.04.2023
|
| Subjects: | |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Parkinson's disease (PD) has a significant impact on people's lives following Alzheimer's, the most common neurological condition. It causes some or all loss of the ability to move, speak, behave, think, and perform other essential tasks. Monitoring the patient's condition on a frequent basis is one of the essential elements of a successful treatment. In order to track the evolution of a condition, physiological analysis results are translated to a metric, the Universal Parkinson's Disease Rating Scale (UPDRS). It is a scoring system that measures the existence and severity of the symptoms. It ranges from 0-176 for patients who are not receiving treatment, with 0 denoting a healthy condition and 176 denoting entire disability. It is divided into three categories: motor activities, everyday living activities, and thought, behavior, and mood. The motor UPDRS measures the functions including speech, facial expression, stiffness, and tremor on a scale of 0 to 108, with 0 signifying normal status and 108 denoting significant movement disorders. Speech spans from 0 to 8, where 8 indicates incoherent speech. Nowadays, in health care environments, noninvasive tele-monitoring is a growing alternative that may allow for reliable, affordable disease screening and relieving the strain of frequent, clinic visits. As a result, the burden on the costs is reduced, and the clinical evaluation of the subject's condition is evaluated more accurately. The information acquired from the patients can be utilized to accurately anticipate how severe a patient's PD symptoms will be. The proposed study uses a variety of models such as linear regression, decision tree, random forest, gradient boost and xgboost with parameter optimization for maximizing the prediction accuracy. The random forest model has the highest R2 value and the lowest RMSE.We also investigate the relationship between the two dependent target variables, total UPDRS and motor UPDRS. The two dependent target variables were found to have a 0.947 correlation coefficient. Therefore, any of the two target variables is enough and be used to train a model. |
|---|---|
| DOI: | 10.1109/ICONSTEM56934.2023.10142432 |