Converting Electron Microscopy Nanoparticle Images to Numerical Data

At the upcoming SciX 2026 Conference, which is set to take place in Sparks, Nevada, Caelin Celani, Postdoctoral Fellow at the University of Delaware, will deliver a presentation on her team’s research. The title of the presentation is “Exploration of Electrodeposited Nanoparticle Synthesis Spaces via Image Analysis and Chemometrics,” and it will discuss a new computational framework that converts electron microscopy images of nanoparticles into standardized numerical data.^1,2

What is the SciX Conference?

SciX is known as "The Scientific eXchange," and traces its roots to 1974.² The conference was founded by the Federation of Analytical Chemistry and Spectroscopy Societies (FACSS). FACSS is a nonprofit federation whose member societies, which include the American Chemical Society (ACS)'s Division of Analytical Chemistry, the American Society for Mass Spectrometry (ASMS), the Society for Applied Spectroscopy (SAS), and the Royal Society of Chemistry (RSC)'s Analytical Division, help shape the conference's scientific program each year.²

What will Celani’s presentation cover?

Celani’s presentation will highlight how her team’s work examined an ongoing issue in materials science, and that is the lack of a quantitative way to describe nanoparticle shape and connect it to the conditions under which particles were made.² To address this issue, Celani and her team applied a novel framework that utilizes chemometric and machine learning (ML) methods to scanning electron microscopy (SEM) images, generating an explicit descriptor matrix that captures nanoparticle morphology in a form suitable for statistical analysis.²

What has been the outcome of not having this descriptor matrix?

As Celani noted in the abstract of her talk, the absence of such a system has kept the structural information contained in microscopy images largely inaccessible to data-driven workflows.² The result is that researchers were then required to depend on qualitative, visual assessments of particle shape when optimizing synthesis methods.² Because nanoparticle morphology directly affects material performance in applications ranging from catalysis to electronics, that gap has limited how systematically scientists can tune synthesis conditions to achieve a desired structure.²

How did researchers test the framework they developed?

Celani and her team tested their framework on two electrochemically deposited nanoparticle systems, copper(I) oxide (Cu₂O) and silver (Ag). By using a five-factor Resolution V fractional factorial design of experiments, the research team systematically varied synthesis conditions across 16 runs for each material and collected SEM images of the resulting particles.² An image analysis pipeline then segmented individual particles and extracted 63 morphological descriptors per particle, quantifying size, shape, and texture.²

What else did Celani’s team investigate in their research?

The research team also examined four supervised classification algorithms against four feature-selection strategies. The four feature-selection strategies included the full descriptor set, a version reduced through principal component analysis, a set based on statistically significant factors from the design of experiments, and a combined approach.² Meanwhile, the four supervised classification algorithms included partial least squares discriminant analysis (PLS-DA), support vector machines (SVMs), random forest (RF), and k-nearest neighbors.²

What were the results of the study?

At SciX, Celani will explain how her team discovered that the Cu₂O particles behaved differently from the Ag particles. Cu₂O particles sorted into three morphological groups that were linearly separable, meaning all four classification algorithms performed comparably well (kappa greater than 0.93).² On the other end, Ag particles formed a continuous range of shapes rather than distinct groups, and only nonlinear methods produced strong classification results, with SVMs reaching a kappa of approximately 0.84.²

References

1. Wetzel, W.; Spectroscopy Staff. Previewing the Upcoming 2026 SciX Conference. Spectroscopy Online, 2026. https://www.spectroscopyonline.com/view/previewing-the-upcoming-2026-scix-conference (accessed September 11, 2026).
2. Celani, C. Exploration of Electrodeposited Nanoparticle Synthesis Spaces via Image Analysis and Chemometrics. Presented at the SciX 2026 Conference, Sparks, Nevada, October 6, 2026. Available at: https://www.scixconference.org/onlineprogram