Related Experiment Video
Updated: Jul 15, 2025
10:28
Dynamic Digital Biomarkers of Motor and Cognitive Function in Parkinson's Disease
Published on: July 24, 2019
15.2K
Radiomics for Parkinson's disease classification using advanced texture-based biomarkers.
Sonal Gore1, Aniket Dhole1, Shrishail Kumbhar1
1Pimpri Chinchwad College of Engineering, Nigdi, Pune, Maharashtra, India.
Methodsx
|October 4, 2023
Summary
This study introduces an advanced texture analysis using Local Binary Patterns (LBP) on MRI scans for Parkinson's disease (PD) detection. The method achieved up to 83.33% accuracy, offering a faster, more precise diagnostic tool for neurodegenerative diseases.
Area of Science:
- Neuroimaging
- Biomedical Engineering
- Radiomics
- Machine Learning
Background:
- Parkinson's disease (PD) diagnosis is typically manual and time-consuming.
- Computer-aided diagnosis using MRI can enhance diagnostic precision and speed.
- Texture-based radiomic analysis offers potential for identifying subtle disease markers.
Purpose of the Study:
- To investigate the efficacy of advanced texture analysis using Local Binary Pattern (LBP) variants for Parkinson's disease classification from MRI scans.
- To develop and evaluate a computer-aided diagnostic model for Parkinson's disease.
- To explore the potential of radiomic features for early and accurate PD detection.
Main Methods:
- Radiomic analysis was performed on 3D T1-weighted and resting-state MRI scans from 72 subjects (36 healthy controls, 47 Parkinson's patients).
- Local Binary Pattern (LBP) method with custom variants was applied to extract textural biomarkers from 360 selected 2D MRI images.
- Feature selection using recursive feature elimination reduced ~150-300 LBP histogram features to 13-21 significant features, analyzed with SVM and random forest algorithms.
Main Results:
- Variant-I of the LBP method achieved the highest test accuracy of 83.33%, with precision of 84.62%, recall of 91.67%, and F1-score of 88%.
- Classification accuracies ranged from 61.11% to 83.33%, with AUC-ROC values between 0.43 and 0.86 across four LBP variants.
- The proposed method, utilizing an SVM classifier with 10-fold cross-validation, demonstrated significant potential in detecting Parkinson's patients.
Conclusions:
- Advanced biomedical texture features, specifically extended LBP variants, can effectively detect subtle variations in local appearance indicative of Parkinson's disease.
- The developed radiomic analysis model shows promise as a precise and rapid computer-aided diagnostic tool for Parkinson's disease.
- Texture-based radiomic analysis of MRI scans represents a valuable approach for improving the diagnostic workflow for neurodegenerative conditions like PD.
