Stacked Model-Based Classification of Parkinson's Disease Patients Using Imaging Biomarker Data

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Parkinson's disease (PSD) diagnosis is challenging due to symptom overlap with other neurodegenerative disorders.
• Clinical assessment relies on history and symptoms, often insufficient for definitive diagnosis.
• Neuroimaging biomarkers offer potential for objective disease assessment.

Purpose of the Study:

• To develop a machine learning model for accurate Parkinson's disease identification.
• To utilize neuroimaging biomarkers to analyze dopamine deficiency in the brain.
• To differentiate Parkinson's disease patients from healthy controls (HC).

Main Methods:

• A stacked machine learning (ML) model was proposed for classification.
• The model integrated predictions from K-nearest neighbor (KNN), random forest (RFA), and Gaussian naive Bayes (GANB) algorithms.
• Neuroimaging biomarker data, specifically dopamine levels, were used for analysis.

Main Results:

• The proposed stacked ML model achieved 92.5% accuracy in identifying Parkinson's disease.
• The model demonstrated superior performance compared to traditional diagnostic schemes.
• Dopamine level patterns effectively indicated disease severity and progression.

Conclusions:

• Stacked machine learning models can accurately diagnose Parkinson's disease using neuroimaging data.
• Neuroimaging biomarkers are valuable tools for assessing Parkinson's disease severity and progression.
• This approach enhances the differential diagnosis of neurodegenerative diseases.