Differentiating idiopathic Parkinson's disease from multiple system atrophy-P using brain MRI-based radiomics: a...

Area of Science:

• Neuroimaging
• Radiomics
• Machine Learning

Background:

• Accurate differentiation between idiopathic Parkinson's disease (IPD) and multiple system atrophy-parkinsonian type (MSA-P) is crucial for patient management.
• These conditions have distinct prognoses and treatment responses, necessitating precise diagnostic tools.

Purpose of the Study:

• To develop and validate a radiomics-based model for distinguishing IPD from MSA-P using MRI data.
• To assess the diagnostic performance of machine learning classifiers applied to radiomic features.

Main Methods:

• A multicenter retrospective study involving 287 patients (186 IPD, 101 MSA-P) who underwent brain MRI.
• Radiomic features were extracted from T1- and T2-weighted MRI sequences.
• Various machine learning models, including SVM, were trained and tested; a nomogram combining clinical and radiomic data was also evaluated.

Main Results:

• The support vector machine (SVM) model, forming the basis of the Rad-signature, demonstrated strong performance (AUC 0.885-0.900).
• The radiomics-based signature significantly outperformed clinical-only models.
• A nomogram integrating radiomic and clinical features achieved the highest diagnostic accuracy (AUC 0.973-0.963) and clinical utility.

Conclusions:

• Radiomics analysis of MRI is a powerful method for differentiating IPD from MSA-P.
• Integrating radiomic and clinical data enhances diagnostic accuracy and supports personalized treatment strategies.
• The developed nomogram shows promise for improving diagnostic workflows in clinical practice.