Optical surface information-based respiratory phase-sorting and motion-incorporated reconstruction for SPECT imaging

Area of Science:

• Medical imaging
• Nuclear medicine
• Image reconstruction

Background:

• Respiratory motion during SPECT imaging causes blurring artifacts, leading to inaccurate activity and absorbed dose estimations.
• Motion artifacts compromise image quality and quantitative accuracy in SPECT scans.

Purpose of the Study:

• To develop and validate an optical surface imaging (OSI) system for respiratory signal extraction in SPECT.
• To implement and evaluate GPU-accelerated motion-incorporated reconstruction algorithms for improved SPECT imaging.
• To produce motion-free SPECT images by integrating accurate respiratory motion information.

Main Methods:

• Simulated SPECT projections using the 4D XCAT Phantom and SIMIND Monte Carlo program.
• OSI system modeled with Gaussian noise; respiratory signals compared to center-of-light (COL) approach.
• Validated OSI with patient data comparing OSI-derived signals to cone-beam CT (CBCT) projections.
• Evaluated 4D reconstruction (4D-Recon) and post-reconstruction registration (post-Recon) against non-motion-corrected (non-MC) and gated images.

Main Results:

• OSI demonstrated high correlation with ground-truth respiratory signals (0.99 ± 0.004), outperforming COL methods.
• Patient data showed high consistency between OSI-based and CBCT-based respiratory signals.
• Both 4D-Recon and post-Recon significantly improved recovery coefficients (RC) and contrast recovery coefficients (CRC) compared to non-MC and Gating.
• 4D-Recon showed advantages over post-Recon in specific quantitative metrics.

Conclusions:

• The novel OSI approach provides accurate and reliable respiratory signal extraction for SPECT.
• Motion-incorporated SPECT reconstruction using OSI significantly enhances image quality and quantitative accuracy.
• This method holds potential for improving absorbed dose quantification in radiopharmaceutical therapy.
• Open-source reconstruction algorithms are available via the PyTomography library.