A Scalable Platform for Real-Time Microscopy Image Analysis Using Artificial Intelligence and Machine Learning

Recent increases in data collection hardware over the past several decades have resulted in the observation that most data-rich microscopy workflows are now software limited. Artificial Intelligence and Machine Learning methods could eliminate many of the software bottlenecks by performing automated data analysis and image evaluation. Given this, the adoption of these advanced methods has been hampered by the lack of combined hardware-software platforms that make their use Ĺturn-keyĺ in common microscopy workflows. Theia Scientific has developed a platform that eliminates the barriers to entry for using Artificial Intelligence and Machine Learning for routine and advanced microscopy activities. The technology leverages a modern web application software stack and Graphical Processing Unit-equipped edge and near-edge computing devices to host and run state-of-the-art Artificial Intelligence and Machine Learning models. Multiple models can be run in parallel by forming ad-hoc heterogenous clusters of multiple Graphical Processing Unit-equipped devices. This scalable technology enables dynamic resource utilization to analyze multiple data streams and disparate acquisition modalities often too complex and information dense for manual interpretation. This multi-model clustering technology for multi-modal microscopy data analysis will be further advanced to provide a new user experience that incorporates model ensemble creation integrated with object tracking algorithms, a novel adaptive learning workflow for on-the-fly model development and training set generation, and user and data management utilities to fully integrate with large multi-instrument user facilities with data isolation and user access policies and requirements. The proposed efforts will then provide a rich user experience within a modern user interface that enables the technology to extend past niche applications previously demonstrated in prior efforts. The resulting outcome will be a flexible and scalable platform that reduces labor costs by up to 80%, decouples expertise in multi-modal microscopy data analysis tasks, and accelerates the innovation lifecycle of new technologies that rely on quantitative digital microscopy with multiple data streams, such as technologies and products under development in material science, biology, medical imaging, geology, energy storage, energy consumption, and environmental monitoring.