Searching scale space for activation in PET images

Area of Science:

• Neuroimaging
• Statistical analysis
• Medical physics

Background:

• Positron Emission Tomography (PET) images of cerebral blood flow (CBF) are typically smoothed, reducing resolution and potentially obscuring activation details.
• Standard smoothing is often fixed (e.g., 20 mm FWHM), and statistical maps are searched for local maxima, which can be suboptimal.

Purpose of the Study:

• To address the challenge of assessing the statistical significance of peaks identified through multi-dimensional smoothing in neuroimaging data.
• To develop a unified P value for 4-D scale-space peaks in Gaussian fields, improving the accuracy of activation region detection.

Main Methods:

• Proposed a 4-D search over smoothing kernel widths and spatial dimensions, extending previous work by Poline and Mazoyer.
• Developed a unified P value calculation for pooled-variance Z-statistic images (Gaussian fields).
• Validated the method for accuracy across regions of varying shapes and sizes.

Main Results:

• A unified P value for 4-D local maxima was derived, offering accurate significance assessment.
• The method is applicable to Gaussian statistical fields, including those from functional Magnetic Resonance Imaging (fMRI).
• Enables estimation of activation region size and location when peaks are well-separated.

Conclusions:

• The developed method provides an accurate and unified approach to assessing the significance of scale-space peaks in neuroimaging.
• This advancement enhances the analysis of PET and fMRI data by improving the detection and characterization of brain activation regions.