Contextual High-Throughput 3D Volume Electron Microscopy Data Acquisition Using Artificial Intelligence

One of the greatest challenges when generating large and/or high-resolution three-dimensional (3D) volume electron microscopy (vEM) datasets is long acquisition times. We developed a method that leverages Artificial intelligence algorithms to increase acquisition throughput for 3D datasets by creating an AI-derived mask for the targeted region of interest, referred to as Adaptive Scanning. This allowed for specific structures to be imaged at high resolution, with the surrounding area captured at lower resolution, without artificially generating or enhancing any raw data. We demonstrate that this dynamic-resolution scanning approach significantly reduced acquisition time depending on the number of pixels of interest and was sample agnostic, being compatible with a diverse array of organisms and tissues, including brains, parasites, cultured cells, and plants. This multiresolution strategy has the potential to enhance focused ion beam scanning electron microscopy and other forms of vEM, by increasing time savings by up to twofold or more, enabling routine generation of multiple and/or larger vEM datasets more efficiently, cost-effectively, and allowing more data collection to increase statistical power for comparative studies.