Near real time tomographic 3D reconstruction with the use of the PC graphics accelerator

In this article the authors describe a new 3D reconstruction system for x-ray tomography. The goal was to provide a surgeon 3D x-ray tomography images online during an operation. The system was designed, for the use with intraoperative, medical x-ray devices (C-arms). It consists of a 3D volume reconstruction algorithm, implemented using a common pc graphics accelerator, to minimize the computation time. Due to mechanical limitations of the x-ray system the algebraic reconstruction algorithm SART was used. The known reconstruction algorithm was enhanced for an iterative online reconstruction, where the computation is done at the same time the x-ray system acquires new images. The surgeon can add more x-ray images to improve the results of the 3D reconstruction. Due to this enhancement and the high speed computation, the surgeon is able to acquire the needed number of x-ray images for a specific clinical case and to avoid unnecessary radiation. For clinically relevant 3D data 30 to 100 x-ray images are needed, depending on the anatomical region and the needed information. The reconstruction system was verified and validated using a mathematically described phantom and human x-ray images from clinical trails. In less than ten seconds after the last image was taken a 3D reconstruction is calculate with the new online reconstruction approach. The results show no significant difference in terms of image quality compared to conventional SART reconstructed images, where the reconstruction time is about 3 minutes.