Low-dose sinogram restoration enabled by conditional GAN with cross-domain regularization in SPECT imaging.

Area of Science:

• Medical Imaging
• Radiological Sciences
• Artificial Intelligence in Healthcare

Background:

• Low-dose single-photon emission computed tomography (SPECT) imaging is crucial for reducing patient radiation exposure.
• Acquiring high-quality SPECT images under low-dose conditions presents challenges due to increased noise and reduced contrast.
• Sinogram denoising is essential for improving the quality of reconstructed SPECT images from low-dose acquisitions.

Purpose of the Study:

• To develop and evaluate a novel sinogram denoising method for low-dose SPECT imaging.
• To suppress random oscillations and enhance contrast in the projection domain of SPECT sinograms.
• To improve the overall quality of reconstructed SPECT images obtained under low-dose acquisition protocols.

Main Methods:

• Proposed a conditional generative adversarial network with cross-domain regularization (CGAN-CDR) for low-dose SPECT sinogram restoration.
• The generator incorporates multiscale sinusoidal feature extraction and long skip connections for enhanced feature reuse and information recovery.
• Utilized a patch discriminator for detailed feature characterization and implemented cross-domain regularization in both projection and image domains.

Main Results:

• CGAN-CDR demonstrated significant noise and artifact suppression, along with enhanced contrast and structure preservation in low-dose SPECT sinograms.
• Visual and quantitative analyses confirmed superior performance in both global and local image quality metrics compared to existing methods.
• The model showed robustness in recovering detailed structures, particularly bone structures, from high-noise sinograms.

Conclusions:

• The CGAN-CDR method is feasible and effective for restoring low-dose SPECT sinograms, yielding substantial quality improvements.
• This approach enables potential applications in real-world low-dose SPECT studies, enhancing diagnostic accuracy while minimizing radiation dose.
• The developed technique offers a promising solution for high-quality SPECT imaging in clinical settings with reduced acquisition times or radiation exposure.