Computer-Aided Classification Framework of Parkinsonian Disorders Using ¹¹C-CFT PET Imaging

Area of Science:

• Neuroimaging
• Artificial Intelligence in Medicine
• Neurology

Background:

• Parkinsonian disorders (PDs) present diagnostic challenges.
• Accurate differential diagnosis is crucial for effective patient management.
• Dopamine transporter (DAT) imaging offers insights into neurodegenerative processes.

Purpose of the Study:

• To evaluate a novel computer-aided classification framework for differentiating PDs.
• To assess the utility of 11C-methyl-N-2β-carbomethoxy-3β-(4-fluorophenyl)-tropanel (11C-CFT) PET imaging in this context.
• To determine the diagnostic performance of the framework across various PD subtypes.

Main Methods:

• Development of a multistep computer-aided classification framework.
• Integration of MRI-assisted PET segmentation, feature extraction, and automatic classification.
• Utilized a random forest method for feature relevance assessment.
• Tested performance using early and advanced disease stages.

Main Results:

• Achieved high accuracy rates for Parkinson's disease (85.0%), MSA (82.2%), and PSP (89.7%).
• Overall accuracy of the framework was 80.4%.
• Caudate and putamen regions showed highest diagnostic relevance; midbrain contribution was minimal.
• Combined early and advanced disease staging improved diagnostic metrics (sensitivity, specificity, PPV, NPV) for most subtypes.

Conclusions:

• The developed computer-aided classification framework shows significant potential for improving PDs differential diagnosis.
• 11C-CFT PET imaging is a valuable tool when integrated with advanced computational analysis.
• This AI-driven approach could enhance diagnostic accuracy and patient care in movement disorders.