Project: NanoTerasu Integration
Sendai is home to NanoTerasu, a 3 GeV high-brilliance synchrotron radiation facility. We build the analysis infrastructure that takes its measurement data beyond being captured, to being understood.
What we work on
One thread is speeding up image reconstruction for X-ray CT: shortening the time from measurement to image, so that results can be checked while the experiment is still running. The other is designing AI that automatically extracts meaningful regions from CT images. Working with a research group in synchrotron imaging, we aim to flag features such as the spots where fracture begins in a material, before they are overlooked.
Why it matters
The resolution of the instruments keeps improving, but reading meaning out of the images still depends on the experience of individual researchers. When that stage becomes the bottleneck, the facility's capability stops translating into results. By connecting measurement with computation, we aim to draw out the full value of the data NanoTerasu produces.
Main activities
- Accelerating image reconstruction for phase-contrast CT
- Semantic extraction from CT images using Transformer self-attention
- Building CT analysis pipelines on large-scale parallel computing
- Exploring Ising formulations of manifold learning for measurement data (NanoTerasu data among the intended targets)
Keywords
Synchrotron radiation / NanoTerasu / X-ray CT / Image reconstruction / Semantic-extraction AI