AI- and ML-driven Multimodal Analysis of Nanoparticle Morphology, Composition, and Dynamics

The shape, composition, and kinetic properties of nanoparticles essentially define their application potential in catalysis, advanced materials, and nanomedicine. Current SEM/TEM analysis techniques rely on laborious manual processing steps that limit throughput and introduce subjectivity, which can potentially discourage large-scale studies. A Segment Anything Model (SAM)-based framework was developed for accurate nanoparticle segmentation, integrating EDS data and graph neural networks (GNNs) trained on time-resolved images to provide comprehensive characterization. The proposed pipeline delivers predictions of catalytic performance that are 15% more accurate than those obtained with conventional regression models, reduces analysis time from days to hours, and remains robust when characterizing diverse nanoparticle morphologies. The approach provides quantitative morphological parameters and elemental distribution patterns, as well as sub-second resolution for tracking aggregation dynamics. Such an integrated methodology enables rapid screening and targeted optimization of nanoparticles, especially for nanoparticle-driven applications such as drug delivery platforms and catalytic systems, where particle shape plays a decisive role in intrinsic performance.