SMGDiff: step mapping generalized diffusion model for efficient noise reduction in cardiac-gated myocardial perfusion...

Area of Science:

• Medical Imaging
• Computational Imaging
• Nuclear Medicine

Background:

• Improving cardiac gating myocardial perfusion single-photon emission computed tomography (CG MP-SPECT) image quality is vital for accurate diagnosis.
• Traditional diffusion models (DM) for MP-SPECT denoising are computationally intensive and time-consuming.
• There is a need for efficient and lightweight generalized diffusion models for CG MP-SPECT image denoising.

Purpose of the Study:

• To develop and evaluate a novel, lightweight generalized diffusion model for efficient denoising of CG MP-SPECT images.
• To reduce the computational resources and processing time required for MP-SPECT image denoising.
• To enhance the diagnostic accuracy of CG MP-SPECT through improved image quality.

Main Methods:

• Proposed a novel step mapping generalized diffusion model (SMGDiff) using cardiac-gated MP-SPECT images as diffusion endpoints and a mean-preserving degradation operator.
• Implemented a stepwise mapping and error optimization module (SMEO) to minimize cumulative errors during reconstruction.
• Evaluated the model on a retrospective dataset of 50 MP-SPECT scans (CG-8 and CG-16), using quantitative metrics (PSNR, SSIM, NMSE) and a reader study.

Main Results:

• SMGDiff-5 achieved superior performance across all metrics, significantly outperforming CNN, U-Net, GAN, and DDPM.
• Demonstrated exceptional computational efficiency, requiring only 0.024 s per slice compared to 4.982 s for a 1000-step DM.
• Reader studies confirmed SMGDiff-5's high image quality and diagnostic confidence, comparable to static MP-SPECT.

Conclusions:

• The SMGDiff-5 model offers robust and efficient denoising for CG MP-SPECT images.
• It provides superior performance over traditional deep learning methods.
• Significantly reduced computational demand makes it a practical solution for clinical applications.