Ultrafast 3D reconstruction for X-ray real-time tomography (RTT)

Imaging complete objects in a tomographic manner in nearly real-time is of interest for many applications. One example is imaging of patients with a very high frame rate to be able to observe and study variations in the patients anatomy (e.g. in the heart or lung) with high temporal resolution. Another example is baggage inspection where the object is transported at speeds as high as 0.5 m/s. Here, reconstruction and subsequent threat detection must be completed before a bag reaches the switch in the baggage belt that redirects suspicious bags. The scope of this paper is a real-time CT system that does not require physically moving parts. This is achieved by using multi- focus x-ray sources and a number of detector rings surrounding the object. We briefly describe the scanner design and then concentrate on the signal processing pipeline of the scanner. Image reconstruction is driven by a cluster of cell blades and a highly performant backprojection algorithm that allows for full vectorization. Thereby, a continuous reconstruction of 240 to 480 images per second, each consisting of 768×768 pixels and receiving contributions from 384 to 768 focal spots, is achieved with a total of four cell processors. While the real-time tomograph is a dedicated luggage scanner the concept realized and validated is of highest interest for clinical CT, too. The possibility of defining arbitrary scanning sequences may yield further improvements in image quality regarding the scanning of moving objects or regarding dose-efficient scans with organ-dependent scan trajectories.