No ground truth needed: unsupervised sinogram inpainting for nanoparticle electron tomography (UsiNet) to correct missing wedges

• DOI: 10.1038/s41524-024-01204-x
• Publication Date: 2024-02-02T12:03:36Z
• Authors: Lehan Yao; Zhiheng Lyu; Jiahui Li; Qian Chen
• Keywords: Inpainting; Ground truth

Abstract Complex natural and synthetic materials, such as subcellular organelles, device architectures in integrated circuits, alloys with microstructural domains, require characterization methods that can investigate the morphology physical properties of these materials three dimensions (3D). Electron tomography has unparalleled (sub-)nm resolution imaging 3D a material, critical for charting relationship among synthesis, morphology, performance. However, electron long suffered from an experimentally unavoidable missing wedge effect, which leads to undesirable sometimes extensive distortion final reconstruction. Here we develop demonstrate Unsupervised Sinogram Inpainting Nanoparticle Tomography (UsiNet) correct wedges. UsiNet is first sinogram inpainting method be realistically used experimental by circumventing need ground truth. We quantify its high performance using simulated nanoparticles (NPs). then apply tomographs, where >100 decahedral NPs vastly different byproduct are simultaneously reconstructed without distortion. The sorted based on their shapes understand growth mechanism. Our work presents potent tool advance tomography, especially heterogeneous samples datasets large wedges, e.g. collected beam sensitive or during temporally-resolved in-situ imaging.