Clustering dynamic PET images on the Gaussian distributed sinogram domain.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Processing

Background:

• Dynamic PET imaging requires accurate kinetic parameter estimation.
• Individual pixel time-activity curves (TACs) suffer from low signal-to-noise ratio (SNR), leading to inaccurate parameter estimates and high spatial variance.
• Clustering pixels into regions and averaging TACs improves SNR.

Purpose of the Study:

• To extend sinogram-domain clustering methods for dynamic PET images.
• To address the challenge of clustering when sinogram data are Gaussian distributed, rather than Poisson distributed.

Main Methods:

• Dynamic PET image clustering was performed in the sinogram domain.
• The method was adapted for Gaussian-distributed sinogram data, accounting for corrections like scatter, randoms, and attenuation.
• Clustering aimed to group pixels with similar kinetic characteristics.

Main Results:

• The developed method allows for clustering of dynamic PET images in the sinogram domain with Gaussian-distributed data.
• This approach enhances the SNR of time-activity curves (TACs) by averaging signals within identified regions.
• Improved SNR facilitates more accurate kinetic parameter estimation.

Conclusions:

• Clustering dynamic PET images in the sinogram domain is a viable and effective preprocessing step.
• The presented method extends sinogram-domain clustering to handle Gaussian-distributed data, broadening its applicability.
• This technique improves the reliability and accuracy of kinetic parameter estimation in dynamic PET studies.