Synthesizing PET images from High-field and Ultra-high-field MR images Using Joint Diffusion Attention Model

TL;DR

The paper tackles the scarcity of PET data and the infeasibility of ultra-high-field PET-MRI by introducing the Joint Diffusion Attention Model (JDAM), which learns the joint distribution between PET and MRI to synthesize PET from MRI. JDAM diffusion-processes the PET modality while keeping MRI fixed and uses a predictor-corrector sampling scheme guided by a U-Net-based score network to realize conditional generation from $p(\mathbf{x},\mathbf{y})$. On the ADNI dataset, JDAM outperforms state-of-the-art CycleGAN for 3T MRI PET synthesis and demonstrates feasibility for ultra-high-field MRI (5T/7T), producing PET images with realistic spatial distributions and improved SNR. This cross-modal diffusion approach holds promise for expanding PET-era insights without additional radiation exposure and may facilitate future ultra-high-field multimodal imaging workflows.

Abstract

MRI and PET are crucial diagnostic tools for brain diseases, as they provide complementary information on brain structure and function. However, PET scanning is costly and involves radioactive exposure, resulting in a lack of PET. Moreover, simultaneous PET and MRI at ultra-high-field are currently hardly infeasible. Ultra-high-field imaging has unquestionably proven valuable in both clinical and academic settings, especially in the field of cognitive neuroimaging. These motivate us to propose a method for synthetic PET from high-filed MRI and ultra-high-field MRI. From a statistical perspective, the joint probability distribution (JPD) is the most direct and fundamental means of portraying the correlation between PET and MRI. This paper proposes a novel joint diffusion attention model which has the joint probability distribution and attention strategy, named JDAM. JDAM has a diffusion process and a sampling process. The diffusion process involves the gradual diffusion of PET to Gaussian noise by adding Gaussian noise, while MRI remains fixed. JPD of MRI and noise-added PET was learned in the diffusion process. The sampling process is a predictor-corrector. PET images were generated from MRI by JPD of MRI and noise-added PET. The predictor is a reverse diffusion process and the corrector is Langevin dynamics. Experimental results on the public Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset demonstrate that the proposed method outperforms state-of-the-art CycleGAN for high-field MRI (3T MRI). Finally, synthetic PET images from the ultra-high-field (5T MRI and 7T MRI) be attempted, providing a possibility for ultra-high-field PET-MRI imaging.

Figures (10)

• Figure 1: Schematic diagram of the joint diffusion attention model.
• Figure 2: Overview of the proposed U-Net architecture incorporating Resnet block and Attention block into the U-Net.
• Figure 3: The architecture of a Resnet block and an Attention block used in the U-Net.
• Figure 4: Comparison of synthetic results in the axial plane. The first row shows MRI, corresponding ground truth PET and synthetic PET results obtained using score-based SDE, CycleGAN, and JDAM methods in the upper layer of the axial plane. The second row shows images whose order is the same as the first row in the middle layer of the axial plane, while the third row shows images whose order is the same as the first row in the lower layer. The yellow numbers in the bottom left corner indicate the PSNR (dB) and SSIM between real PET and synthesis PET.
• Figure 5: Comparison of synthetic results in the coronal plane. The first row shows MRI, corresponding ground truth PET and synthetic PET results obtained using score-based SDE, CycleGAN, and JDAM methods in the upper layer of the coronal plane. The second row shows images whose order is the same as the first row in the middle layer of the coronal plane, while the third row shows images whose order is the same as the first row in the lower layer. The yellow numbers in the bottom left corner indicate the PSNR (dB) and SSIM between real PET and synthesis PET.