Image processing platform

OpenXi is an accessible hardware and software platform composed mostly of common, off the shelf components and open-source software (where appropriate).

The OpenXi platform can scale from simple, laptop-resident implementations, through mid-range systems capable of several TeraFLOPS dedicated to image processing, all the way to a large, clustered, image processing server.

OpenXi leverages the latest progress in GPU and FPGA-based computing to achieve many times the performance which would be needed to implement a typical DSA (Digital Subtraction Angiography) image processing chain, all the way from the detector correction, to the final DSA display, while leaving plenty of computing power available for new algorithms.

X-ray machine interfaces

OpenXi works with X-ray equipment vendors and detector and C-arm system manufacturers to develop the necessary interfaces to gain  access to their real-time X-ray image data (including some related signals, like time-correlated ECG etc.). The physical interlace modules, which we call the "ImageTaps", assure  minimum intrusion to the existing systems and regulatory compliance through the use of such techniques as passive optical splitters and other fail-safe interfaces.

This collaboration is based on the  mutual interest to accelerate progress of real-time X-ray image processing and development of related applications.

ImageTaps will support several fiber-optic interfaces and protocols which are in use today such as Aurora, GigaLink and Fiber Channel. We also plan to support the GigaBit Ethernet and several parallel buses,  RaceWay, and similar - depending on the demand.

To protect the proprietary information of the X-ray equipment vendors supporting our project, the video stream between the ImageTap modules and the processing platform can be encrypted. Decryption to an open form will be performed by the platform after applying those vendor-specific algorithms (i.e. detector correction) which x-ray equipment manufacture wish to keep secret.

Real-time image processing library

This is a hardware accelerated  image processing library for the  the Python language. To achieve tight real-time performance with minimum latency, it provides the  mechanism of the "PixelStream", which handles video transport from the ImageTap through the image processing pipeline. The library and framework also supports off-line image processing and testing the hardware-accelerated algorithms against their versions written in the classic, non-accelerated, Python.

Connectivity

Initially connectivity to the image acquisition module is via proprietary fiber-optic interface, but it will be migrated to optical USB 3.0 as soon as this standard becomes available.

Sharing the results and off-line applications are supported by standard DICOM interfaces.