Abstract
Statistical Parametric Mapping (SPM) is a computational approach for analysing functional brain images like Positron Emission Tomography (PET). When performing SPM analysis for different patient populations, brain PET template images representing population-specific brain morphometry and metabolism features are helpful. However, most currently available brain PET templates were constructed using the Caucasian data. To enrich the family of publicly available brain PET templates, we created Chinese-specific template images based on 116 [<sup>18</sup>F]-fluorodeoxyglucose ([<sup>18</sup>F]-FDG) PET images of normal participants. These images were warped into a common averaged space, in which the mean and standard deviation templates were both computed. We also developed the SPM analysis programmes to facilitate easy use of the templates. Our templates were validated through the SPM analysis of Alzheimer's and Parkinson's patient images. The resultant SPM t-maps accurately depicted the disease-related brain regions with abnormal [<sup>18</sup>F]-FDG uptake, proving the templates' effectiveness in brain function impairment analysis.
Introduction
In neuroimaging, researchers frequently employ Statistical Parametric Mapping (SPM) to identify brain regions exhibiting abnormal metabolic activity. This process requires a patient’s brain scan to be "warped" or registered to a standard reference image, known as a template. Historically, the most widely used templates, such as the MNI-305 or ICBM-152, have been constructed from Caucasian datasets.
However, human brain morphology varies significantly across different ethnic populations. When a Chinese subject's scan is registered to a Caucasian template, it can lead to significant anatomical distortion. This mismatch introduces noise and potential errors when quantifying metabolic changes in neurodegenerative diseases like Alzheimer’s or Parkinson’s. To address this, we developed the first open-access, Chinese-specific brain PET template based on [18F]-FDG imaging.
Methodology
To create both a mean template (representing average metabolism) and a standard deviation template (representing normal variation), we implemented a rigorous two-step pipeline:
- Spatial Normalization: 116 subject images were mapped into the "Chinese2020" MR atlas space using Symmetric Normalization (SyN). This non-rigid registration algorithm ensures that the warping is smooth and maintains anatomical realism.
- Intensity Normalization: Pixel intensities were normalized by dividing the image intensity by the mean value of pixels within the 40–90% range of maximum brain voxel intensity.
- Data Re-filtering: To ensure the integrity of the "normal" group, a "three-sigma rule" based on the Sum of Squared Differences (SSD) was applied. This resulted in the removal of 3 outliers from the initial 119 participants.
To facilitate widespread adoption, we also developed Matlab-based SPM analysis programs that allow these templates to be easily integrated into existing research workflows.
Results and Validation
The effectiveness of the Chinese-specific template was demonstrated through several key metrics and clinical validations:
- Reduced Anatomical Distortion: Analysis of the Jacobian determinant (|J|) showed that our template resulted in values much closer to 1.0 (isometric) compared to Caucasian templates, which required extreme warping (ranging from 0.4 to 1.8) to accommodate Chinese brain structures.
- Clinical Accuracy: The templates were tested on patients with Alzheimer’s Disease (AD) and Parkinson’s Disease (PD). The resulting SPM t-maps accurately identified classic biomarkers: reduced metabolism in the temporal and parietal lobes for AD, and abnormalities in the basal ganglia for PD.
- Voxel-wise Variance: By providing a standard deviation template, researchers can now calculate voxel-wise z-scores, offering a more nuanced view of how a patient's metabolism differs from the population's natural variance. Notably, the frontal region showed relatively larger metabolic variances among healthy participants.
Discussion and Conclusion
While this study utilized a robust sample of 116 participants, it is limited by being a single-center retrospective dataset. Future work should aim to incorporate multi-center data and develop templates for other specific PET tracers, such as those for amyloid or tau imaging.
By minimizing inter-population anatomical differences, these Chinese-specific templates significantly improve the accuracy of statistical mapping for neurodegenerative diseases in Asian populations. The templates and associated software have been made publicly available on NITRC and GitHub to support the global neuroimaging community.
Summary updated