Sparse Sampling Electron Microscopy Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Scanning electron microscopes (SEMs) face limitations in speed and efficiency due to traditional raster scanning methods, which result in lengthy data collection times and are challenging in high noise environments, especially when trying to capture detailed structures in materials science and neuroscience applications.
Innovation Solution
Implementing a sparse sampling methodology where an electron beam or probe scans a subset of randomly or pseudo-randomly designated pixel locations, combined with a compressive sensing inversion method using a total variation prior, to reconstruct images efficiently, allowing for faster data collection and image reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional raster scanning is used to visit every pixel location, then complete image coverage is achieved, but data collection time becomes excessively long
Solution Approach 1:
The patent extracts only a subset of pixel locations (randomly or pseudo-randomly selected) from the complete image grid, visiting only these sampled locations rather than every pixel. This extraction approach reduces data collection time while the compressive sensing reconstruction algorithm recovers the complete image from these sparse measurements, achieving speedups of 4× or more compared to full raster scanning.
Solution Approach 2:
The patent applies partial action by collecting data from only a fraction of the total pixel locations (e.g., 25% or less of all pixels). Instead of performing complete raster scanning, the electron beam visits only a partial set of strategically random locations, and the reconstruction algorithm fills in the missing information, dramatically reducing acquisition time while maintaining image quality.
2Productivity
If sparse sampling is used to reduce data collection time, then imaging speed increases, but image quality and detail preservation deteriorate
Solution Approach 1:
The patent introduces a compressive sensing reconstruction algorithm with total variation prior as an intermediary between the sparse measurements and the final image. This intermediary processing step recovers the complete image from the undersampled data by exploiting the smoothness and structure of natural images, preserving image quality and details even when sampling density is reduced to 25% or less of full raster scanning.
Solution Approach 2:
The patent changes the sampling parameter from systematic raster scanning to random or pseudo-random sampling patterns. This parameter change in the sampling strategy, combined with the total variation prior in reconstruction, allows the system to achieve high imaging speed while maintaining image quality by capturing essential image information more efficiently than traditional methods.
3Loss of time
If random sampling is used to speed up acquisition, then data collection is faster, but structured details in the sample may be missed
Solution Approach 1:
The patent applies preliminary action by using a total variation prior in the reconstruction algorithm that pre-establishes expectations about image smoothness and structure. This prior knowledge is incorporated before seeing the data, allowing the reconstruction to correctly interpret random samples and recover structured details that would otherwise be missed, preventing information loss while maintaining fast acquisition speeds.
Data Source
AI summary
Systems and methods for conducting electron or scanning probe microscopy are provided herein. In a general embodiment, the systems and methods for conducting electron or scanning probe microscopy with an undersampled data set include: driving an electron beam or probe to scan across a sample and visit a subset of pixel locations of the sample that are randomly or pseudo-randomly designated; determining actual pixel locations on the sample that are visited by the electron beam or probe; and processing data collected by detectors from the visits of the electron beam or probe at the actual pixel locations and recovering a reconstructed image of the sample.