Parkinson's disease classification using gait analysis via deterministic learning.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Machine Learning

Background:

• Gait analysis is crucial for assessing mobility and detecting motor deficits in Parkinson's disease (PD).
• Quantitative gait data can aid in the diagnosis and management of PD.
• Current methods may require further refinement for precise PD detection.

Purpose of the Study:

• To develop and validate a novel method for classifying Parkinson's disease (PD) patients and healthy controls using gait analysis.
• To leverage deterministic learning theory and radial basis function (RBF) neural networks for gait pattern classification.
• To assess the diagnostic performance of the proposed method in distinguishing between PD and healthy gait patterns.

Main Methods:

• Gait dynamics were extracted from vertical ground reaction forces during usual and self-selected paces.
• Radial basis function (RBF) neural networks were employed to approximate gait dynamics for training data.
• A classification phase utilized dynamical estimators and error metrics based on deterministic learning theory.

Main Results:

• The proposed method achieved high classification accuracy (96.39%), sensitivity (96.77%), and specificity (95.89%) in distinguishing PD patients from healthy controls.
• Five-fold cross-validation was used to evaluate the performance on a dataset of 93 PD patients and 73 healthy controls.
• The study demonstrated the effectiveness of RBF networks and deterministic learning in gait-based PD classification.

Conclusions:

• The developed gait analysis method, based on deterministic learning theory and RBF networks, effectively differentiates between Parkinson's disease patients and healthy individuals.
• The identified gait features and classification approach show significant potential for clinical application in PD diagnosis.
• This research contributes to the advancement of non-invasive diagnostic tools for neurodegenerative movement disorders.