Diffusion transformer model with compact prior for low-dose PET reconstruction.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Positron emission tomography (PET) is vital for non-invasive clinical diagnosis.
• Low-dose PET scans reduce radiation exposure but yield insufficient data for high-quality image reconstruction.
• Image quality is critical for accurate diagnosis and reliable outcomes.

Purpose of the Study:

• To enhance the reconstruction quality of low-dose PET imaging using a novel diffusion transformer model (DTM).
• To address challenges in low-dose PET imaging, including noise and loss of critical details.
• To improve diagnostic accuracy and patient safety through advanced imaging techniques.

Main Methods:

• Proposed a diffusion transformer model (DTM) for joint optimization of diffusion and transformer models.
• Integrated a lesion refining block and alternating direction method of multipliers (ADMM) for improved lesion recovery and detail preservation.
• Utilized compact prior guidance within the DTM framework.

Main Results:

• DTM achieved state-of-the-art performance in low-dose PET image reconstruction across multiple metrics (PSNR, SSIM, NRMSE, CR, COV).
• Demonstrated superior denoising and lesion preservation capabilities compared to baseline methods, even at ultra-low-dose levels (1%).
• Showcased robust generalization performance on both phantom and patient datasets.

Conclusions:

• The DTM effectively reconstructs high-quality low-dose PET images, reducing radiation exposure.
• This approach ensures reliable imaging for early disease detection and clinical decision-making.
• DTM offers a promising tool for advancing clinical and research applications in PET imaging.