Design of an Optimized Feature Driven Severity Stage Classifier for Parkinson’s Disease Prediction Using Deep Learning
Parkinson’s Disease (PD) is a degenerative neurological condition that seriously affects both motor and non-motor abilities. Among many biomarkers, speech abnormalities have become a potential sign of PD as the disease affects the motor control of the vocal point. Therefore, speech data i...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11122434/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Parkinson’s Disease (PD) is a degenerative neurological condition that seriously affects both motor and non-motor abilities. Among many biomarkers, speech abnormalities have become a potential sign of PD as the disease affects the motor control of the vocal point. Therefore, speech data is a convenient, non-invasive, and valuable diagnostic tool. This work presents a novel methodology for classifying the severity stage of parkinson’s disease using speech data, addressing a notable deficiency in existing research works that have focused on binary classification. Stage classification distinguishes PD’s early, mid, and late phases, offering essential insights for stage-specific therapy strategies. We proposed a Parkinson’s Speech Stage-Net model (PSS-Net) for the prediction of the severity of PD that includes Isolation Forest-based outlier detection and removal, KMeansSMOTE data balancing, and Optimized Dimensionality Reduction techniques. The essential speech features were found using a tree-based feature selection method. Selected features are fed into the Optimized and Fused Neural Network model. The proposed model effectively classified the stages of PD with an accuracy of 89%. Moreover, this research study stresses openness using SHAP (SHapley Additive exPlanations) to understand feature contributions, thereby improving the clinical relevance and credibility of the model. |
|---|---|
| ISSN: | 2169-3536 |