Electron microscopes are powerful tools that allow researchers to study nanoscale objects and processes with high spatial and temporal resolution. To achieve this high resolution, the microscope’s electron beam bombards specimens with large doses of electrons, which can damage the specimen. Any adjustments to the microscope can increase the electron dose, exacerbating the problem. In addition, an electron microscope generates data in the form of images at a rate of more than 1,000,000 pixels per second, creating a burden on computer processing and storage.

Scientists at Pacific Northwest National Laboratory developed a faster, more efficient, automated approach to producing images from an electron microscope. The patented approach subsamples the specimen using an electron beam probe. It then generates sparse datasets that can be used to estimate specimen characteristics and reconstruct an image with high accuracy. Characteristics include feature dimension, composition, and frequency count of one or more features. This type of approach—paired with computational imaging techniques, such as compressive sensing—has shown great promise in minimizing specimen damage and analysis time.

In the past, reconstructing an image from subsampling was a time-consuming process that had to be done manually. Full image analysis might take anywhere from tens of minutes to entire weeks. PNNL’s automated approach allows estimation of specimen characteristics from sparse data quickly and simply, in near-real-time. The result is an image that is substantially equivalent to one produced from fully sampling the specimen but is obtained more efficiently and with minimal electron-beam damage to the specimen. 

APPLICABILITY

The approach can be used with any electron microscope to observe and measure parameters of specimens, such as human cells, plant tissues, and complex environmental samples. It can also be used to identify when an image/process has significantly changed from previous images (i.e., event detection).