Non-stationary spatial filtering and accelerated curve fitting for parametric imaging with dynamic PET

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Biophysics

Background:

• Dynamic Positron Emission Tomography (PET) imaging is crucial for assessing physiological parameters.
• Conventional image processing methods often struggle with noise and contrast, limiting quantitative accuracy.
• Accurate quantification of physiological parameters is essential for diagnosing and monitoring diseases.

Purpose of the Study:

• To develop and implement an advanced image processing technique for dynamic PET studies.
• To improve image quality by reducing noise while preserving essential contrast.
• To enhance the quantitative accuracy of physiological parameter estimation in dynamic PET.

Main Methods:

• Development and implementation of a non-stationary spatial low pass filter.
• Integration with an accelerated non-linear curve fitting routine.
• Application to 18F-2-fluoro-2-deoxyglucose (FDG) dynamic PET studies.

Main Results:

• Generated high-contrast, low-noise images of physiological parameters including blood volume, kinetic rate constants, precursor pool volume, and glucose metabolism.
• Demonstrated significant noise reduction and contrast preservation in both simulated (pie phantom) and real patient data (brain infarct).
• Observed considerable improvement in the quantitative accuracy of pixel parameter values compared to unprocessed or conventionally smoothed images.

Conclusions:

• The developed non-stationary spatial low pass filter combined with accelerated non-linear curve fitting significantly improves image quality in dynamic PET.
• This advanced processing method enhances the reliability and accuracy of quantitative analysis of physiological parameters.
• The technique shows great potential for improved diagnostic capabilities in various clinical applications, particularly in neurology.