CT scan image reconstruction

At the core of any CT scan image reconstruction

is a computer algorithm called Filtered Back projection (FBP). Each of the hundreds of x-ray image data sets obtained by the CT scanner is filtered to prepare them for the back projection step. Back projection is nothing more than adding each filtered x-ray image data set’s contribution into each pixel of the final image

reconstruction. Each x-ray view data set consists of hundreds of floating point numbers, and there are hundreds of these data sets. In a high-resolution image, there are millions to tens of millions of pixels. It is easy to see why summing hundreds of large data sets into millions of pixels is a very time-intensive operation which only gets worse as the image resolution increases.

     Image reconstruction implementation

SRC’s IMPLICIT+EXPLICIT™ Architecture is well suited to accelerating CT scan image reconstruction. In the simplest SRC-7 system implementation, a microprocessor is paired with a Series H MAP® processor. The system microprocessor provides data input and displays the final image using a commodity graphics card. The MAP processor contains an instantiation of the FBP algorithm. These two processors working together achieve a 29x performance boost over the 3.0 gigahertz 64-bit Xeon microprocessor working alone.