CUDA Spotlight: Tools for Microsurgeons




Compute The Cure

Find other great interviews in our CUDA Spotlights Archive.

This week's Spotlight is on Kang Zhang, a PhD candidate in Electrical and Computer Engineering at Johns Hopkins University. His interests include GPU-accelerated biomedical imaging.

NVIDIA: Kang, what are you working on at Johns Hopkins?

Kang: My current research focuses on interventional Optical Coherence Tomography (OCT) technology for microsurgery.

Conventionally, visualization during microsurgery is realized with a surgical microscope, which limits the surgeon’s field of view and causes limited depth perception of micro-structures and tissue planes beneath the surface.

Such issues commonly exist in many kinds of microsurgeries such as ophthalmic surgery, neurological surgery and otolaryngologic surgery.

OCT is a new imaging modality capable of non-invasive 3D micrometer-resolution imaging, which makes it highly suitable for guiding microsurgery. As part of my PhD work, I developed an ultra-high-speed, real-time OCT imaging system using a hardware-software platform based on GPU technology. 

NVIDIA: How does GPU computing play a role in your work? 
Kang: For a clinical interventional imaging system, real-time imaging acquisition, reconstruction and visualization are all essential. However, in current OCT technology, the reconstruction and visualization speeds are generally far behind the data acquisition.

Therefore, current high-speed OCT systems are usually non-realtime and working in a “post-processing mode,” which limits its interventional application. To solve these bottlenecks, I use CUDA technology to accelerate both the OCT image reconstruction and visualization, in particular, 3D volume rendering.

NVIDIA: How did you become interested in this area? 
Kang: When I tried to solve the image processing bottlenecks for my interventional OCT project, I looked at various high-performance computing methods including multi-core cluster, FPGA and GPGPU. After comparing computing performance, cost and system integration, I chose CUDA as the ultimate solution.

The CUDA acceleration is highly cost-effective compared to the overall cost of an OCT system and no optical modification is required. I am also working on developing a CUDA-based high data throughput imaging platform for general purpose applications.

NVIDIA: What are some advantages of working with CUDA? 
Kang: Thanks to CUDA’s great parallel processing ability, we achieved a >20X speedup of OCT image reconstruction and demonstrated the first GPU-based real-time 4D (3D + time) OCT system. Several GPU based algorithms have been developed to further improve image quality such as non-uniform fast Fourier transform and numerical dispersion compensation, which was time consuming on CPUs and now can be running in real-time on GPUs.

 
4D Optical Coherence Tomography Imaging 
Demo of GPU-based real-time 4D OCT technology, providing comprehensive spatial view of micro-manipulation region with accurate depth perception. Image reconstruction performed by NVIDIA GTX 580 and volume rendering by NVIDIA GTS 450. The images are volume rendered from the same 3D data set. Imaging speed is 5 volumes per second. Each volume has 256×100×1024 voxels, corresponding to a physical volume of 3.5mm×3.5mm×3mm.

NVIDIA: What are the potential real-world applications? 
Kang: CUDA-accelerated real-time OCT technology opens the way for interventional OCT imaging for microsurgery, which can potentially enhance the surgeon’s visibility and instrument manipulation, diminish surgical risk, improve surgical outcomes and give significant promotion to the present state-of-the-art microsurgery technologies.

NVIDIA: As computing becomes faster, what can we look forward to?
Kang: With increasing GPU computing capability, it will be feasible to transfer more and more computationally-intensive tasks from huge computer clusters right onto our desktops, with much less power consumption and lower cost. More GPUs will be embedded in existing computer systems as compact, cost-effective and green supercomputing engines.

Kang Zhang’s Biography:
Kang Zhang is currently a Ph. D. Candidate in Electrical and Computer Engineering at Johns Hopkins University. His research interests include GPU-accelerated biomedical imaging, high data throughput imaging platform, real-time 4D imaging system, and optical sensor based medical devices. From 2009 to 2010, Kang worked as an ORISE Research Fellow for the U. S. Food and drug Administration (FDA), where he developed optical metrology methods for medical device evaluation.

For more information, see: