Reference-Free Isotropic 3D EM Reconstruction using Diffusion Models
This work addresses a domain-specific challenge in electron microscopy analysis, offering a novel solution for researchers in biomedical imaging and related fields.
The paper tackles the problem of anisotropic axial resolution in electron microscopy images by proposing a diffusion-model-based framework for isotropic 3D reconstruction without needing reference data or prior knowledge about degradation, achieving robustness and superiority over supervised methods as demonstrated on two public datasets.
Electron microscopy (EM) images exhibit anisotropic axial resolution due to the characteristics inherent to the imaging modality, presenting challenges in analysis and downstream tasks.In this paper, we propose a diffusion-model-based framework that overcomes the limitations of requiring reference data or prior knowledge about the degradation process. Our approach utilizes 2D diffusion models to consistently reconstruct 3D volumes and is well-suited for highly downsampled data. Extensive experiments conducted on two public datasets demonstrate the robustness and superiority of leveraging the generative prior compared to supervised learning methods. Additionally, we demonstrate our method's feasibility for self-supervised reconstruction, which can restore a single anisotropic volume without any training data.