Electron Tomography Artifact Correction for Wide-Field Isotropic Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electron tomography methods face challenges in achieving both a wide field of view and isotropic resolution, particularly when samples are prepared as slides, due to limited angular ranges leading to anisotropic resolution and missing areas in Fourier space.
Innovation Solution
A machine learning model, such as a convolutional artificial neural network or generative adversarial neural network, is trained to correct artifacts in electron tomography reconstructions by learning from a complete angular range, allowing it to generate artifact-free reconstructions with improved resolution from incomplete angular range acquisitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a sample is prepared as a slide to achieve a wider field of view, then the field of view is improved, but the angular range is limited leading to anisotropic resolution
Solution Approach 1:
The patent applies preliminary action by training an AI model in advance using complete angular range data. The model learns to recognize and correct artifacts before actual tomography reconstruction. When processing limited angular range data, the pre-trained model can compensate for missing information and correct anisotropic resolution artifacts without requiring actual complete angular range acquisitions.
Solution Approach 2:
The patent introduces an AI model as an intermediary between the limited angular range projection data and the final tomography reconstruction. This intermediary model processes the incomplete data, fills missing information, and produces artifact-corrected reconstructions with improved isotropic resolution, effectively mediating between the constrained input data and the desired output quality.
2Manufacturing precision
If a sample is prepared as a point to achieve complete angular range [-90°; 90°], then the resolution isotropy is improved, but the field of view is limited
Solution Approach 1:
The patent uses copying by training the AI model on complete angular range data (from point-prepared samples) to create a digital model that can generate complete reconstruction information. This trained model then copies the artifact-correction capabilities to process slide-prepared sample data, effectively transferring the resolution isotropy benefit from point-preparation to slide-preparation workflows.
3Manufacturing precision
If dual-axis tomography is used to reduce missing angular area, then the reconstruction quality is improved, but the electron dose and acquisition time are doubled
Solution Approach 1:
The patent replaces the mechanical dual-axis tomography approach with an AI-based computational method. Instead of physically rotating the sample holder 90° and performing a second acquisition series, the AI model computationally compensates for the missing angular information. This substitution eliminates the need for doubled electron dose and extended acquisition time while achieving similar or superior reconstruction quality.
4Area of stationary object
If a limited angular range is used for slide preparation, then the field of view is wide, but missing areas in Fourier space cause anisotropic resolution
Solution Approach 1:
The patent applies self-service by enabling the AI model to autonomously identify and fill missing Fourier space information. The model processes the limited angular range data, detects gaps in Fourier space coverage, and self-corrects by generating appropriate information based on learned patterns from complete angular range training data, without requiring external intervention or additional acquisitions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The model effectively corrects artifacts in electron tomography reconstructions, enabling samples prepared as slides to achieve isotropic resolution and a wider field of view, reducing blurring and distortion.
Implementation Method 1
using a transmission electron microscope capable of emitting an electron beam towards the sample
Data Source
Figure 1a~1b
Figure 2
Figure 3
AI summary
Method for training an automatic artifact correction model in an image of a sample obtained by electron tomography, comprising the steps of: - Performing (201) an acquisition of a first set of projection data of a sample using a transmission electron microscope capable of emitting an electron beam towards the sample, for a plurality of inclination angles of the sample relative to the electron beam belonging to a first angular range, - Determining (202) a first set of tomographic reconstructions of the sample from the first set of projection data, - Determining (203) a second set of degraded tomographic reconstructions of the sample from a subset of the first set of projection data associated with a subset of inclination angles of the first angular range,- Train (204) the artifact correction model to correct the second set of degraded tomographic reconstructions in order to obtain the first set of tomographic reconstructions.