NANOPARTICLE ANALYSIS BY MACHINE LEARNING

Synthesis of high-quality nanoparticles for sensing, biolabelling and diagnostic imaging requires careful control over, not just the mean characteristics, but the statistical distributions of particle characteristics. Statistical distributions can be generated using AFM, SEM and/or TEM to image large numbers of nanoparticles, to obtain information on size, shape, structure and composition. However, manual analysis of such images is time consuming, and is limited to a fairly small number of particles, which provides only limited accuracy.

We offer advanced software for the automated analysis of TEM, SEM or AFM images containing large numbers of nanoparticles. Generated are accurate statistical distributions of the size, shape, relative affinity and orientation of potentially millions of nanoparticles in a given sample. Automation is based on unsupervised machine learning, requiring no training data, and only minimal user interaction. Our algorithm offers the unique advantage of accommodating, not just highly dispersed particles, but also moderately agglomerated particles.