Objective: Spatial normalization of C-Pittsburgh Compound B (PiB) images is challenging for an automatic quantitative analysis without magnetic resonance imaging (MRI) because of different distribution patterns between amyloid positive and negative images. To overcome this issue, we evaluated a selective positron emission tomography template (SPT) method.
Materials and methods: Three sets of single positron emission tomography templates were created: PiB negative template, PiB positive template, and mixed template. Sixty-one patients with dementia were enrolled as the validation cohort. Magnetic resonance imaging-aided normalization method was used as a reference. The SPT method was based on visual classification (positive, negative, and equivocal). The optimal templates for each visual group were determined by correlation values and average percent errors (APEs) with MRI-aided normalization. The results of the SPT and the single template methods were compared with those of MRI-aided normalization in terms of correlation values, APEs, and concordance rates.
Results: The SPT (PiB negative template for the negative and equivocal groups and PiB positive template for the positive group) showed higher correlations and concordance rate and lower APEs with MRI-aided normalization than did the single template.
Conclusions: Use of the SPT provides accurate normalization of amyloid images without MRI.