Image Reconstruction Technology Utilizes GPUs to Speed Up CT Scanner Workflows

Advanced image reconstruction technology has recently been developed for the medical imaging, security, and nondestructive testing markets.

Image reconstruction for computed tomography (CT) scanners frequently takes hours of processing time and represents a major obstacle to efficiency for many organizations. The AxRecon system reduces the time required to complete this work from hours to minutes to speed up imaging workflow. Acceleware Corp. (Calgary,AB,Canada), a leading developer of acceleration solution for high-performance computing, developed the system. AxRecon was recently presented at the European Congress of Radiology (ECR) in March 2008, held in Vienna, Austria.

"Our entry into the medical and commercial imaging market space represents a significant growth opportunity for Acceleware, and is consistent with Acceleware's long-term growth strategy to tackle new markets,” said Sean Krakiwsky, CEO of Acceleware. "The benefits of higher throughput and image quality that are already being realized by early customers of AxRecon will help us drive adoption of this product within the broader imaging market.”

The long processing time required for CT image reconstruction limits throughput and quality for companies, clinicians, and researchers. To meet the demands of their workload, these users frequently forfeit image quality to save on processing time. AxRecon works with existing cone beam CT scanners to eliminate the bottleneck by significantly speeding up filtered back projection computations. This enables medical and commercial users to accelerate the reconstruction of their data and improve the quality of their images without disrupting their current workflow.

The AxRecon system combines Acceleware's proprietary software acceleration libraries with the massive parallel processing power of NVIDIA (Santa Clara, CA, USA) graphic processing units (GPUs) to dramatically increase the speed of sophisticated computations. AxRecon integrates with existing CT scanners with minimal setup. By accelerating algorithms such as filtered back-projection image reconstruction, clients are able to increase the speed of their operations and quality of their end results, while reducing the energy use and physical footprint of their high performance computing systems.

"In biomedical micro-computed tomography applications, the volume dimensions of our reconstructions have increased dramatically over the past few years,” said David Holdsworth of Robarts Imaging. "During the same period, scan times have also been reduced; this makes cone-beam CT reconstruction time the bottleneck that limits workflow in some situations. GPU-based reconstruction is a cost-effective solution for this task, and we have found that Acceleware's AxRecon product can provide 3D [three-dimensional] reconstructions up to 50 times faster than a single-CPU [central processing unit].”

Acceleware develops and markets systems that enable software vendors to leverage heterogeneous, multi-core hardware environments without having to rewrite their applications for parallel processing. This acceleration abstraction technology results in end-users achieving significant computing speed-ups and migrating vendor applications from single-core processing to multiple-core CPUs, GPUs or other acceleration products.

Acceleware's third-generation multi-board GPU solutions can accelerate simulation and processing algorithms by over 35 times, reducing data processing from multiple hours to minutes.


Related Links:
Acceleware