Cognitive activity analysis of Parkinson's patients using artificial intelligence techniques.

Area of Science:

• Artificial Intelligence in Medicine
• Neurological Disorder Diagnostics
• Machine Learning Applications

Background:

• Parkinson's disease (PD) diagnosis relies on identifying characteristic movement disorders and tremors.
• Early detection of PD is crucial for effective management and treatment.
• AI-driven models offer promising avenues for improving diagnostic accuracy and efficiency.

Purpose of the Study:

• To investigate the efficacy of AI models in detecting Parkinson's disease using hand-drawn spiral data.
• To evaluate different machine learning algorithms for classifying drawings from PD patients and healthy controls.
• To explore the potential of hand drawing analysis as a non-invasive, decision-support tool for clinicians.

Main Methods:

• A dataset was created with 40 PD patients and 40 healthy controls (HC) drawing spirals on a tablet.
• Support Vector Machine (SVM), Random Forest (RF), and Naive Bayes (NB) classifiers were employed for distinction.
• Data preprocessing involved min-max normalization, and Leave-One-Subject-Out (LOSO) Cross-Validation (CV) was used to prevent overfitting.
• Principal Component Analysis (PCA) was applied for dimension reduction to enhance classifier performance.

Main Results:

• The highest classification accuracy achieved was 90% using the SVM classifier.
• Optimal performance was observed with non-template drawings combined with PCA.
• The study demonstrated the potential of SVM with PCA for PD detection from drawings.

Conclusions:

• Hand drawing analysis, utilizing simple gestures, can serve as a non-invasive pre-evaluation system in clinical settings.
• This AI-based approach minimizes confounding factors like environmental and educational differences.
• The findings pave the way for hand drawing analysis as an auxiliary system to aid healthcare professionals in PD diagnosis, saving valuable time.