The Evolution of AI from Image Interpretation toward Scientific Inference in Nanoparticle Electron Microscopy

This review surveys how AI is advancing nanoparticle electron microscopy across TEM, HRTEM, STEM, and in situ TEM. It covers particle detection, segmentation, morphology quantification, atomic-resolution restoration, defect identification, 2D-to-3D structural inference, and dynamic-process analysis. The authors organize methods from conventional machine learning and CNNs to transformers, self-supervised learning, foundation models, multimodal AI, and physics-informed learning. The review also examines integration with simulations, metadata, autonomous experimentation, benchmarking, and data requirements, framing microscopy as a platform for scientific inference and accelerated materials discovery.