Dr. Matthew Andrew directs the development of microscopy technologies and workflows for the geosciences at Carl Zeiss Microscopy. His research interests focus on flow and transport processes in porous media, including multiscale digital core analysis, in situ analysis of multiphase flow, unconventional pore and mineral systems, and ways to automate big-data analyses using modern data science, particularly using automated optical petrography. His work also includes the development and implementation of new machine learning techniques to improve imaging speed, enhance resolution, and enable unique workflows within ZEISS Microscopy and the development of customized image processing and analysis workflows in collaboration with academic and industrial partners.
Before working at Carl Zeiss Microscopy, Dr. Andrew worked at Imperial College developing the first reservoir condition in situ flooding rig for an X-ray microscope, enabling fluid flow to be directly visualized for the first time at representative subsurface conditions.
Machine learning can be embedded deeply into image acquisition systems, removing noise and artifacts from images, making them cleaner, sharper, quicker to acquire, and easier to interpret. We will introduce ZEISS DeepRecon for X-ray microscopy, allowing acquisition times to be reduced by up to a factor of 10 for repetitive samples.
These technologies can also be embedded into integrated analysis-acquisition workflows where areas for high resolution analysis can be automatically selected from low resolution overview images, dramatically reducing the requirement for user operation and increasing system automation. As these workflows tend to be highly targeted at the specific analyses being performed and often require customization, we will overview some new ZEISS SolutionLab capabilities, where ZEISS and users work together in partnership to develop specific solutions to challenging problems.
Finally, we will overview how AI can be embedded within analytical workflow to allow for analyses that were previously impossible to perform. This could include upscaling localized analyses over much larger areas, estimating 3D effective properties from rapidly acquired 2D data, or even propagating localized detailed 2D analytics through 3D data through a combination of correlative microscopy and advanced machine learning.