Super-resolution in brain positron emission tomography using a real-time motion capture system

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Standard positron emission tomography (PET) imaging is limited by spatial resolution.
• Motion artifacts significantly degrade PET image quality, hindering the visualization of fine anatomical details.
• Accurate motion measurement is crucial for advanced image reconstruction techniques.

Purpose of the Study:

• To develop and evaluate a super-resolution (SR) estimation framework for brain PET.
• To leverage high-resolution infrared tracking for precise motion measurement in PET.
• To improve the spatial resolution and visualization capabilities of brain PET imaging.

Main Methods:

• Developed a temporal and spatial calibration for integrating an infrared tracking camera (NDI Polaris Vega) with a PET/CT scanner.
• Implemented a list-mode Ordered Subset Expectation Maximization (OSEM) PET reconstruction algorithm incorporating real-time motion correction.
• Acquired data using moving phantoms and non-human primate (NHP) models on a GE Discovery MI PET/CT scanner.

Main Results:

• Super-resolution reconstruction yielded visibly increased spatial resolution compared to standard static PET acquisitions.
• Improved visualization of small anatomical structures was achieved in both phantom and NHP studies.
• Quantitative analyses (SSIM, CNR, line profiles) validated the enhanced resolution and accuracy of the SR method.

Conclusions:

• Super-resolution imaging is feasible in brain PET by employing real-time motion tracking with high-resolution infrared cameras.
• This SR framework significantly enhances the diagnostic potential of PET by improving image clarity.
• The integration of optical motion tracking offers a robust solution for motion correction in dynamic PET studies.