CRIS: Cross-Plane Self-Supervised Isotropic Restoration for Anisotropic Volumetric Imaging Across Modalities
Researchers have introduced CRIS, a novel self-supervised framework designed to restore isotropic resolution in anisotropic volumetric medical imaging. This method treats 3D restoration as a 2D stripe completion task, training on synthetically degraded in-plane slices. CRIS has demonstrated superior performance across various MRI and electron microscopy datasets, outperforming existing interpolation and restoration techniques in metrics like PSNR, SSIM, and segmentation consistency. The framework's modality flexibility and lack of requirement for paired isotropic targets make it a versatile tool for improving downstream analysis in volumetric imaging. AI
IMPACT This self-supervised restoration technique could enhance the quality and utility of medical imaging data, potentially improving diagnostic accuracy and research outcomes.