A novel feature extraction method based on dynamic handwriting for Parkinson's disease detection

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Parkinson's disease (PD) is a prevalent neurodegenerative disorder affecting the elderly.
• Handwriting analysis presents a non-invasive and accessible method for PD detection.
• Existing handwriting-based PD detection methods often lack comprehensive feature representation.

Purpose of the Study:

• To develop an advanced feature extraction technique for improved Parkinson's disease detection using handwriting.
• To address limitations in current methods by integrating kinematic, pressure, and angle dynamic features.
• To enhance classification performance through a novel optimization algorithm.

Main Methods:

• A novel moment feature was proposed, integrating kinematic, pressure, and angle dynamic handwriting features.
• Time-frequency-based statistical (TF-ST) features were extracted from dynamic handwriting data.
• An escape Coati Optimization Algorithm (eCOA) was employed for global optimization to improve classification accuracy.

Main Results:

• The proposed method achieved high accuracy, reaching up to 98.67% on tested datasets.
• Excellent sensitivity (average 98.15%) and specificity (average 99.17%) were demonstrated.
• High Area Under the Curve (AUC) values (average 98.66%) indicate robust diagnostic performance.

Conclusions:

• The proposed handwriting feature extraction method significantly enhances Parkinson's disease detection accuracy.
• The integration of diverse dynamic features and advanced optimization offers a promising avenue for PD diagnosis.
• The developed approach provides a sensitive, specific, and accurate tool for identifying Parkinson's disease.