2 min read

Q&A: AI image analysis with Image-Pro webinar

Mar 23, 2026 10:55:32 AM

The most interesting part of a technical webinar often comes after the main presentation in the questions from the audience. During our recent webinar on AI-based image analysis using Image-Pro with Hitachi SEMs, participants raised a series of practical questions about training models, handling noisy images, working with heterogeneous materials, and integrating analysis into everyday SEM workflows.

Individually, these questions touched on topics such as segmentation accuracy, local AI training, real- time analysis, and data export. Taken together, they reveal something important: users are no longer just curious about AI, they want to understand how it fits into real laboratory work.

How much noise in an SEM image can AI-based segmentation tolerate?

Modern machine learning models are surprisingly robust to noise, especially if the training images include realistic variations in contrast and signal quality. In practice, the best results come from training the model using representative SEM images that include the types of noise and variation expected in real measurements.

Does the AI training happen locally on my computer?

Yes. Training and analysis are performed locally in Image-Pro, meaning your data stays on your system. This is particularly important for many industrial and research users who work with sensitive or proprietary data.

Can AI segmentation handle heterogeneous materials?

Yes. One of the strengths of AI-based segmentation is its ability to recognize complex patterns in heterogeneous microstructures. With proper training examples, the model can distinguish multiple phases, particles, pores, or features even when contrast is variable.

Are pre-trained models available for nanoscale particles?

Pre-trained models can provide a useful starting point, but the best accuracy usually comes from training the model on your own images. This ensures the segmentation reflects the specific contrast, resolution, and morphology of your samples.

Can the analysis be used during imaging to guide SEM acquisition?

In many workflows, rapid image analysis can be used to evaluate features and guide imaging decisions. This can help users quickly determine whether sufficient data has been acquired or whether additional imaging is needed. Using EM-Flow, the Hitachi SEM automation tool, it is possible to automate image acquisition, send data for analysis to Image-Pro and return results that can be used to make decisions on next steps. For example identify an area of interest, move there and acquire images at higher magnification.

Is there a trial version of Image-Pro available?

Yes. Users interested in testing the workflow can request a trial version to evaluate AI-based image analysis on their own datasets.

Final Takeaway

Taken together, these questions reflect a shift in how microscopy data is being used.

AI-based image analysis is moving from an experimental concept to a practical tool that helps SEM users extract reliable, quantitative information from complex images. As workflows mature and software becomes easier to use, the focus is no longer on whether AI can work, but on how it can be applied effectively in everyday laboratory practice.

Watch the webinar on demand

Watch Now

Back