Method and device for training particle image restoration model of cryoelectron microscope
By constructing a cryo-electron microscopy particle image restoration model and utilizing deep learning technology and high-quality two-dimensional particle projection image label data, the quality of cryo-electron microscopy single particle images was improved, solving the problem of low image quality and improving the reconstruction accuracy of the three-dimensional structural model.
Patent Information
- Application Number
- CN202410430420.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-10
- Publication Date
- 2025-10-17
AI Technical Summary
The low image quality of single-particle images from cryo-electron microscopy leads to low accuracy in 3D structural models. Existing technologies struggle to effectively improve the quality of single-particle images, especially due to the lack of ideal high-quality images as supervisory labels, resulting in poor image restoration effects.
By constructing a cryo-electron microscopy particle image restoration model, a three-dimensional structural model is built using multiple cryo-electron microscopy single-particle images. Spatial orientation projection and stitching are performed to generate a high-quality two-dimensional particle projection image, which is then used as label data to train a deep learning model to restore cryo-electron microscopy single-particle images.
It improves the clarity and accuracy of cryo-electron microscopy single-particle images, enhances the reconstruction accuracy of three-dimensional structural models, and achieves better image restoration results.
Smart Images

Figure CN120808059A_ABST