Data driven surrogate signal extraction for dynamic PET using selective PCA: time windows versus the combination of...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Image Processing

Background:

• Respiratory motion correction is crucial in positron emission tomography (PET) to reduce artifacts and improve quantitative accuracy.
• Current data-driven methods, like principal component analysis (PCA), are limited to static PET due to tracer kinetics in dynamic acquisitions.
• Existing PCA-based methods are adversely affected by tracer kinetics, restricting their use in dynamic PET imaging.

Purpose of the Study:

• To extend principal component analysis (PCA)-based data-driven motion correction methods for applicability to dynamic PET imaging.
• To develop and evaluate novel approaches for extracting respiratory surrogate signals from dynamic PET data.
• To enable advanced post-acquisition motion correction techniques for dynamic PET studies.

Main Methods:

• Exploration of a moving window approach, similar to Kinetic Respiratory Gating.
• Development of a method to extrapolate principal components from later to earlier time points.
• Implementation of a technique to score, select, and combine multiple respiratory components for improved signal extraction.

Main Results:

• All developed methods yielded superior surrogate signals compared to conventional PCA on dynamic data, showing higher correlation with a gold standard respiratory trace.
• Extrapolation of late time point principal components demonstrated more promising results than the moving window approach.
• The method involving scoring, selecting, and combining components provided the most significant benefits over other tested approaches.

Conclusions:

• This work successfully enables the extraction of accurate respiratory surrogate signals from dynamic PET data earlier in the acquisition process.
• The developed methods overcome limitations of traditional PCA, making data-driven motion correction viable for dynamic PET imaging.
• The improved surrogate signal extraction opens possibilities for applying previously incompatible methods, such as advanced respiratory motion correction, to dynamic PET data.