Step-and-shoot respiratory or cardiac gated images are readily available from many Cone beam geometry microCT scanners on the market today.  While admittedly the simplest method to acquire multiphase image data, step-and-shoot scan methods can spend upwards of 50% of the scan time simply accelerating and decelerating gantry equipment.  A free-running acquisition mode, on the other hand, accelerates the gantry only at the beginning and end of the scan, so less time is wasted.  One downside of this free-running mode, however,  is that acquired raw image data needs to be sorted according to phase prior to image reconstruction, and there's little or no guarantee that images will be acquired at precisely the correct phase at all.  Based on recently published methods, we've been experimenting with a post-scan retrospective gating method implemented on the GE Locus Ultra preclinical scanner, and the image above demonstrates an example result:  10 phases were reconstructed automatically using our GPU reconstruction engine, running on our custom workstation.  The total scan acquisition time was 50 seconds, during which time the CT scanner made 10 full rotations.  The total reconstruction time was 86 seconds.  While our initial goal is to produce a turn-key third-party solution for the Ultra scanner, there's no reason why this can't be extended to other hardware platforms - we'll be investigating other scanners, which don't currently have such gating options, in the upcoming months.

Today we turn our attention to the GE Locus Ultra platform in our continuing set of posts on GPU reconstruction:  Integrating a faster GPU reconstruction engine into the GE Locus Ultra requires more careful consideration than some other CT scanners because of the system's complexity and the need for end-to-end integration.  Without this, the scanner cannot maintain it's incredible scan workflow. Keeping this in mind, we've sought out a solution that improves not only the raw reconstruction speed, but the data access rate to the operator console too.  We've blended the original 32-bit-only console software with a 64-bit host computer that runs GPU reconstructions as well as a 64-bit MicroView for visualization purposes.  The workflow is almost indistinguishable from the original, but is faster in almost all aspects, since data is no longer situated on a remote computer, but rather at the operator's console computer.  See the video below for an introduction to the work we're doing in this area.

Thanks to our friends at Endra Inc., we've had an opportunity to involve ourselves in GPU-based algorithm acceleration. This has ended up, for us, being a substantial deep-dive into technologies such as CUDA and OpenCL, but with significant results. The advantages of GPU acceleration over conventional computing are well known, even in the face of impressive improvements in generalized CPU multi-core capabilities in recent years. These advantages are especially true in the medical imaging field, where almost all compute-intensive tasks in the imaging pipeline are embarrassingly parallel. CT reconstruction is one example that we've chosen to focus on recently - we've started by implementing a flexible GPU-accelerated CT reconstruction framework, that can act as both a platform for algorithm design and development, as well as a de facto replacement for legacy CPU-based reconstruction engines. At the moment, it is capable of reconstructing scan data from GE's Locus, eXplore and Ultra CT products - depending on reception, we may choose to port it to more systems. Our reference implementation uses the well-known stock Feldcamp conebeam backprojection algorithm and is written in python for maximum flexibility. It uses pycuda for the heavy lifting. See below reconstruction results from a variety of platforms -- of particular interest, perhaps, is the NVidia GTX 580 results, for sheer speed, and the NVidia GTX 460M, which is running on a laptop. The latter results compare favorably against a 9-computer beowulf cluster, showing just how far GPU technology has come in recent years. The results are preliminary, and will be updated frequently as our recon engine improves and is tested more widely.

(Note: It seems that the interactive graph does not show up, if you are viewing with Internet Explorer. We apologize for that, but for now, you will have to use one of the other browsers, if you wish to see the results).