Symmetry-Aware Brain MRI Inpainting Using Denoising Diffusion Models

Image inpainting is an essential tool for restoring missing or corrupted regions in brain MRI scans, especially for tumor-affected brains, where methods designed for healthy anatomy can fail. This study introduces a novel symmetry-aware generative inpainting framework based on denoising diffusion probabilistic models (DDPMs) that leverages the pseudo-symmetry of the brain to enhance reconstruction quality and training efficiency. By leveraging the pseudo-symmetry of the human brain, aligning MRI data to the MNI space and incorporating mirrored symmetric contextual information as additional inputs, our approach enhances reconstruction quality and generalization across datasets. Experimental results demonstrate that our method outperforms baseline models, achieving superior fidelity with fewer training iterations. These findings highlight the potential of symmetry-aware inpainting to improve brain MRI reconstruction in clinical applications. The code is publicly available at github.com/AlejandroSantorum/symmetry-mri-inpainting .