Topological Descriptors for Parkinson's Disease Classification and Regression Analysis

Area of Science:

• Computational neuroscience
• Medical informatics
• Data science

Background:

• Current neurological disease diagnosis relies heavily on in-person assessments and qualitative data analysis.
• Automated, stable, and accurate diagnostic methods are needed to streamline Parkinson's disease (PD) evaluation.
• Improved diagnostics can provide patients and families more time for timely interventions.

Purpose of the Study:

• To develop and validate an automated method for Parkinson's disease classification and severity assessment.
• To leverage Topological Data Analysis (TDA) and machine learning for objective PD diagnosis.
• To analyze postural shifts data for improved diagnostic accuracy.

Main Methods:

• The proposed methodology integrates TDA with machine learning tools.
• Parkinson's disease postural shifts data is analyzed using persistence images, a TDA representation.
• The method is applied to a dataset including healthy-elderly, healthy-young, and Parkinson's disease patients.

Main Results:

• The study demonstrates the potential of TDA in classifying healthy individuals from PD patients.
• The method shows promise in diagnosing the severity of Parkinson's disease.
• Topological features are found to be invariant to minor data perturbations, enhancing robustness.

Conclusions:

• Topological Data Analysis offers a robust and accurate approach for automated Parkinson's disease diagnosis and severity assessment.
• Integrating TDA with machine learning can significantly enhance the efficiency and objectivity of neurological disorder diagnostics.
• The developed methodology provides a foundation for future advancements in computational diagnostics for neurodegenerative diseases.