Earlier prediction of Parkinson's disease using cross non-decimated wavelet transform and machine learning algorithm

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder impacting motor function, presenting diagnostic challenges.
• Early detection of PD is crucial for timely intervention and management.
• Voice analysis offers a non-invasive approach for PD assessment.

Purpose of the Study:

• To investigate the efficacy of Cross-Non-Decimated Wavelet Transform (CNDWT) combined with Bayesian Optimized Multiple Linear Regression (BOMLR) for early Parkinson's disease detection.
• To develop and validate a novel signal processing technique for PD identification using voice data.

Main Methods:

• Amplitude-sliced and augmented PD voice data processed using Cross-Non-Decimated Wavelet Transform (CNDWT) with Haar and Daubechies (DB3) wavelets.
• Identification of influential PD data attributes through correlation analysis.
• Application of Bayesian Optimized Multiple Linear Regression (BOMLR) for prediction based on identified attributes.

Main Results:

• The proposed CNDWT-BOMLR method achieved a 99% accuracy in predicting Parkinson's disease.
• Comparison with existing methods demonstrated superior performance of the CNDWT approach.
• The study utilized a dataset comprising 31 voice recordings (23 PD patients).

Conclusions:

• The CNDWT technique, integrated with BOMLR, shows significant promise for accurate and early Parkinson's disease detection.
• This approach offers a highly effective, non-invasive method for PD diagnosis using voice biomarkers.
• Further research can explore larger datasets and diverse populations to confirm generalizability.