Local Pattern Transformation Based Feature Extraction for Recognition of Parkinson's Disease Based on Gait Signals

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Medical Diagnostics

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder characterized by the loss of dopamine-producing neurons.
• Early detection of PD is crucial for managing disease progression and improving patient outcomes.
• Gait analysis offers a non-invasive method for assessing neurological conditions like PD.

Purpose of the Study:

• To evaluate the efficacy of novel feature extraction techniques for Parkinson's disease detection using gait signals.
• To introduce and validate the Symmetrically Weighted Local Neighbour Gradient Pattern (SWLNGP) method for gait analysis in PD patients.
• To compare the performance of SWLNGP against other established feature extraction methods.

Main Methods:

• Human gait signals were analyzed using Local Binary Pattern (LBP), Local Gradient Pattern (LGP), Local Neighbour Descriptive Pattern (LNDP), and Local Neighbour Gradient Pattern (LNGP) for feature extraction.
• Statistical features were derived, and the Kruskal-Wallis test was employed for optimal feature set selection.
• An Artificial Neural Network (ANN) was utilized for classification based on the selected features.

Main Results:

• The Symmetrically Weighted Local Neighbour Gradient Pattern (SWLNGP) method demonstrated superior performance in recognizing Parkinsonian gait.
• SWLNGP achieved an accuracy of 96.28%, sensitivity of 95.57%, and specificity of 95.94%.
• The study confirmed the effectiveness of the selected feature set and ANN classifier for PD gait recognition.

Conclusions:

• SWLNGP is a highly effective feature extraction technique for the recognition of Parkinsonian gait.
• The findings suggest that SWLNGP can significantly contribute to the early and accurate diagnosis of Parkinson's disease.
• This approach holds promise for developing advanced diagnostic tools for neurodegenerative disorders based on gait analysis.