Differentiating Parkinsonian Syndromes Using Distinctive Brain Iron Accumulation Patterns in Susceptibility Weighted Image (SWI)

Recent studies report iron overload in atypical Parkinsonian syndromes. Each Parkinsonian syndromes have different topographic patterns of iron distribution in deep brain nuclei, affected by underlying disease pathologies. In this study, we propose a novel framework that automatically analyzes the disease-specific patterns of iron accumulation using susceptibility weighted image (SWI). To improve the segmentation of deep brain nuclei, we use a combination of both T1-weighted MRI and SWI. To analyze the regional iron heterogeneity with segmentation results, radiomic features are extracted from SWI. Subtypes of Parkinsonian syndromes can be differentiated through SVM classifier that has learned the selected disease representative features.