當Deep Learning遇見3D X-ray顯微鏡 

The physics of X-ray imaging can be complex, so ZEISS XRM researchers studied user habits, dove into their challenges, and employed human-centered design (HCD) principles to enable even the newest user in a busy environment to be immediately productive. NavX™, the new user interface for ZEISS Xradia 630 Versa, guides users through automated workflows with intelligent system insights and delivers experimental results more easily and efficiently while also allowing experienced users to explore the full versatility of the platform.

NavX enables you to automate your workflow and provides guidance on the impact the parameters you've chosen will have on your setup. That guidance is directly embedded in the software, taking you through choices in a natural and familiar way.

Additionally, the NavX File Transfer Utility (FTU) takes the data that is being produced by the microscope and automatically transfers it to other locations so that users have their data where they need it, when they need it. These advancements make NavX much more capable for remote operation, advancing user productivity.

NavX intuitive navigation follows the evolution of the XRM user base and revolutionizes X-ray navigation and control with seamless and integrated workflows complementing the planning and execution of advanced correlative workflows..

The first commercially available deep learning reconstruction technology enables you to increase throughput by up to 10× without sacrificing novel resolution at a distance (RaaD). Alternatively, keep the same number of projections and enhance the image quality further. ZEISS DeepRecon provides significant AI-driven speed or image quality improvement.​

​ZEISS DeepRecon Pro is applicable to both unique samples as well as semi-repetitive and repetitive workflows. Self-train new machine learning network models on-site with an extremely easy-to-use interface. The one-click workflow of ZEISS DeepRecon Pro eliminates the need for a machine learning expert and can be seamlessly operated by even a novice user.​

​ZEISS DeepRecon Pro is now available on ZEISS Xradia Ultra nanoscale XRM.

Mouse lung imaged with Xradia Versa. Sample is iodine stained and captured with 3001 projections.  Reconstruction done using DeepRecon (right). Compared with the equivalent image reconstructed using FDK (left)

MARS is a reconstruction algorithm that is aware of the constituents within a reconstruction. A challenge in X-ray reconstruction in a lab setting is that imaging with a polychromatic source creates different X-ray energies to generate a phenomenon called beam hardening. This effect is particularly challenging when your material is very dense and embedded in relatively less dense material. MARS tells the reconstruction system how to compensate for the effect of extreme beam hardening in the regions between very dense objects. This is important in applications like biomaterials, where you might be looking at implants next to bone or tissue. Or electronics where extremely dense solder balls appear next to other less dense materials on a printed circuit board, generating strong artifacts. MARS reconstructs your images to compensate for these effects. ​

Biomedical metal implant in bone. Without MARS, left. With MARS, right.​

A fast and efficient algorithm-based technology that delivers iterative reconstruction from your desktop, allowing you to achieve up to 4× faster scan times or enhanced image quality with equivalent throughput. ZEISS OptiRecon is an economical solution offering superior interior tomography or throughput on a broad class of samples.​

Camera module: 1200 projections in 90 minutes using standard FDK, left, vs. 300 projections in 22 minutes with OptiRecon, right. Comparable image quality in a fraction of the time.