Structural parameters are superior to eigenvector centrality in detecting progressive supranuclear palsy with machine...

Area of Science:

• Neuroimaging
• Neurology

Background:

• Progressive supranuclear palsy (PSP) is an atypical Parkinsonian syndrome.
• Initial symptoms include falls and impaired eye movement.

Purpose of the Study:

• To identify structural and functional brain alterations specific to PSP using multimodal imaging.
• To compare the diagnostic accuracy of structural versus functional imaging markers.

Main Methods:

• Multi-center study using T1-weighted and resting-state functional MRI.
• Analysis of gray/white matter volume and eigenvector centrality.
• Application of multivariate pattern recognition for patient classification.

Main Results:

• Significant gray/white matter volume reduction in the midbrain, cerebellum, and cerebellar peduncles.
• Cortical thinning observed in the cingulate cortex, medial and temporal gyri, and insula.
• Machine learning models based on structural MRI achieved up to 98% accuracy in classifying PSP patients, especially when focusing on the midbrain.

Conclusions:

• PSP is associated with widespread multimodal brain changes beyond midbrain atrophy.
• Structural brain alterations are more effective than eigenvector centrality for machine learning-based PSP prediction.