Multimodal Feature-Guided Diffusion Model for Low-Dose PET Image Denoising with MRI

Motivation: With the advent of integrated PET/MR imaging devices, the high-resolution anatomical information from MRI can help enhance the quality of PET images. Goal(s): Utilizing MRI structural information to optimize the quality of low-dose PET imaging, thereby reducing the risk of radiation exposure for patients. Approach: We use a cross-modal guided restoration network to fully exploit the modality-specific features of LPET and MR images and employ cross-attention mechanisms and positional encoding at multiple feature levels for better feature fusion. Results: The images generated by our proposed network showed superior performance compared to those produced by other networks in both qualitative and quantitative evaluations. Impact: The research can enhance the quality of low-dose PET imaging and reduce the radiation risk for patients, and it also inspires more feasible imaging solutions for the global health field, holding significant scientific and clinical importance.