Preprocessing of ¹⁸F-DMFP-PET Data Based on Hidden Markov Random Fields and the Gaussian Distribution

Area of Science:

• Neuroimaging
• Medical Physics
• Computational Neuroscience

Background:

• 18F-DMFP-PET imaging visualizes dopamine D2/3 receptors for Parkinson's disease (PD) diagnosis.
• While striatal uptake is dominant, other brain regions in PET scans offer diagnostic value.
• Accurate delimitation of regions of interest is vital for automated PD diagnosis.

Purpose of the Study:

• To introduce a novel preprocessing methodology for 18F-DMFP-PET data.
• To enhance the accuracy of computer-aided diagnosis systems for PD.
• To improve the classification of idiopathic and non-idiopathic PD.

Main Methods:

• Image segmentation using a Hidden Markov Random Field algorithm.
• Normalization of segmented maps to achieve Gaussian histogram distributions.
• Classification of PD subtypes using a Support Vector Machine classifier.

Main Results:

• The proposed method divides neuroimages into 4 distinct maps based on voxel intensity and neighborhood.
• Individual normalization of maps ensures consistent histogram modeling across all neuroimages.
• Preprocessing with the novel method yielded higher accuracy in PD classification compared to existing approaches.

Conclusions:

• The developed preprocessing technique significantly improves the accuracy of automated PD diagnosis using 18F-DMFP-PET data.
• This methodology offers a promising advancement for computer-aided diagnostic tools in neurology.
• The approach facilitates more precise differentiation between PD subtypes.