The purpose of QUAITOM is to establish a data science infrastructure for AI-based quantitativeanalysis of X-ray and neutron tomography (μCT). Applying novel concepts, the power of machine learning will for the first time be made available for the very large and scientifically diverse user community within μCT imaging including biomedical and preclinical science.
μCT data is challenging to analyze due to images being large. Using deep learning for segmenting and quantifying structures in μCT images, you will typically need to train a segmentation model for each dataset, which is very time-consuming. The idea in QUAITOM is to combine the research fields
in μCT imaging and machine learning and utilize the mechanisms that drive research in machine learning. If we provide benchmark data of high quality, that reflects the variability and complexity of typical μCT data, we will be able to get the machine learning community to work on solving problems relevant for μCT imaging community.
QUAITOM will be a research infrastructure in the form of a computational that will provide data to machine learning, provide machine learning-based analysis tools for image analysis, and provide problems and tools for students that wish to specialize in machine learning for μCT imaging.
The timing is perfect for creating this synergy between μCT imaging and machine learning. The world’s first 4th generation synchrotron, MAX IV, and the European Spallation Source, ESS in Lund, Sweden, will generate extremely large 3D tomographic images from start of 2022. Moreover, since 2018 DTU, KU, LU, and MAX IV have joined forces on data analysis and user support in the “Center
for Quantitative Analysis for MAX IV data”, QIM. We will build upon the experience and network obtained in the QIM work for establishing the QUAITOM infrastructure.