Case 04. Image x Informatics
Image Informatics
Informatics is a powerful method for R&D, but until now it could not handle unstructured data such as images. GeXeL makes it possible to convert images into features and use image data as explanatory or objective variables.
Traditional Challenges
In the field of materials development, images (microscope photos, cross-sectional images, in-line camera images, etc.) have often been treated as qualitative information judged 'by eye,' making evaluation dependent on individuals. Yet, as shown in batteries and catalysts, causal relationships often exist between microstructure and material properties, making quantification of images essential.
In recent years, AI-driven materials exploration (informatics) has advanced rapidly, and demand is growing for quantifying images as numerical data (features)[1]. On the ground, however, images and experimental data (formulation, temperature, time, etc.) remain unstructured, and data preprocessing is still a significant burden for informatics.
How Our Technology Solves It
Our GeXeL automatically analyzes particles, bubbles, and tissues from SEM/TEM/optical images using AI, instantly extracting dozens of geometric parameters such as equivalent circular diameter, aspect ratio, circularity, orientation, and aggregation ratio. The extracted features can be used directly as explanatory variables for informatics, and data can be shaped together with non-image experimental data such as formulation and equipment conditions in a seamless workflow.
This enables Image Informatics (ImI)—quantifying images for informatics—and integrates the cycle of feature extraction → data shaping → machine learning → experimentation into R&D. Combined with Bayesian optimization for process condition exploration, complex objectives such as 'maximize the longest particle length while keeping variation (CV) below a specified value' become achievable using image-derived features as objective variables.
Because such condition optimization can be performed with just a few clicks, GeXeL strongly supports the launch of informatics initiatives.
Interactive analysis flow
Image upload
Upload microscopy or process-camera images directly. No special preprocessing is required — GeXeL accepts raw images as-is.
Click a step card to see its details
Industry Adoption
Industrial applications span powders (pigments, fillers, catalyst supports), electrode microstructures in batteries and capacitors, additive dispersion in metals and resins, bubbles/defects in plastic molding and additive manufacturing, and particle dispersion in pharmaceuticals and cosmetics—any domain handling images from the nano- and submicron- to the visible scale.
By basing work on image features, organizations can standardize quality indicators, streamline technical communication across departments, and ensure evidence/traceability for external reviews and certifications. In the future, we aim to link GeXeL features to high-throughput and combinatorial experiments, accelerating development through a cycle of automated planning—experiment—analysis—next plan.
GeXeL contributes to a paradigm shift in R&D and, by supporting the birth of new products, weaves new value into society.
References
- [1] Zhang L, Shao S. Image-based machine learning for materials science. Journal of Applied Physics. 2022;132(10):100701.
Feel Free to Contact Us
Want to solve challenges with GeXeL's image analysis?