Smart microscopy is a new imaging approach that involves rapid imaging, prediction of important subregions, then selective re-imaging. This approach has been validated in reducing imaging beam time in electron microscopy connectomics, but the speedup depends on various imaging workflow parameters. Here we present the first runtime analysis of traditional vs. smart microscopy and show how these parameters can magnify, or diminish potential time savings. We provide a GUI application that calculates the theoretical time savings of smart microscopy from user input parameters describing their imaging workflow. Finally, we measure end-to-end runtime of SmartEM acquisition on an electron microscope to demonstrate two strategies for faster acquisition: mixed-precision neural networks and parallelization of microscope and support computer operations.
Athey et al. (2025) studied this question.