There are various imaging technologies available, including the emerging field of naked-eye 3D displays. This technology allows for the display of three-dimensional images without the need for specialized glasses. While initially used primarily in entertainment, this technology is increasingly finding practical applications in numerous professional fields.
With the advancement of technology, Professor Takeyoshi Dohi and his colleagues at the Department of Mechanical Informatics, Graduate School of Information Science and Technology, University of Tokyo, Japan, have conducted research on NVDIA’s CUDA™ parallel computing platform. They believe that naked-eye 3D display technology has the potential to significantly enhance and promote medical imaging. The next phase of development for this technology is expected to be in the medical field.
Introduction to naked eye 3D display
Naked eye 3D display, also known as Glasses-free 3D imaging or naked-eye stereoscopic imaging, was first developed by a research team at the University of Tokyo in 2000. This technology allows for the creation of real-time, three-dimensional images of cross-sectional views of living bodies obtained through CT or MRI scanning. The resulting volume texture can be shaped into a stereoscopic video display for use in intravenous systems. This technology has revolutionized real-time, stereoscopic, and in-vivo imaging, but also faces certain technical barriers.
During the imaging process, the computer must handle a significant amount of pressure due to the real-time, fast, and subtle nature of the process. The first step is to generate a subject image, which requires detailed refinement through the scanner to obtain a complete plan. The next step is stereoscopic imaging, which involves processing each frame of the image and displaying it from multiple angles simultaneously. The total computational load of naked-eye 3D display is the product of the number of frames and the total number of frames in the video. To achieve real-time display, the entire process must be completed instantaneously.
In a 2001 study, the research team attempted to process and analyze a 512 x 512 image using a Pentium III 800 MHz PC. However, they were unable to achieve the desired effect. It took 12 seconds to create a frame from the subject image to the 3D display. To address this issue, the team re-experimented using the UltraSPARC III 900 MHz, which was the latest machine available at the time, with 60 CPUs, and the most powerful computer available. Despite this, the team was only able to achieve a rate of five frames per second. From a practical perspective, this speed is clearly insufficient for medical applications.
Breakthrough in naked eye 3D display technology
The researchers initially developed a prototype system utilizing the advanced generation GPU, GeForce® 8800 GTX. Upon conducting experiments utilizing the dataset from the 2001 study, it was discovered that performance increased to approximately 13 to 14 frames per second. The research team was astonished to find that the UltraSPARC system, which costs tens of millions of yen and hundreds of times more than a GPU, only delivered roughly three times the performance. Additionally, the team’s research indicates that NVIDIA’s GPUs are at least 70 times quicker than the latest multi-core CPUs. Moreover, the research indicates that for large-scale volume texture data, GPUs provide even more pronounced performance benefits.
Currently, the research team is optimizing the current IV system for Tesla using CUDA, leveraging Tesla™ D870, NVIDIA’s newest desktop supercomputer. This development is expected to lead to even more significant performance improvements.
Moreover, the potential of faster new-generation GPUs can be harnessed without the need to modify existing systems. If debugging large CUDA programs can be achieved in a given environment, CUDA can become a more robust parallel computing development tool, potentially leading to wider adoption in medical imaging processing.
Real-time stereoscopic viewing of CT and MRI images enables doctors to examine tissue states and make diagnoses without resorting to invasive procedures such as biopsies and surgeries. Additionally, such images can be viewed simultaneously by multiple doctors, facilitating communication and collaboration. This allows for some doctors to perform arthroscopic surgery and other minimally invasive surgical techniques simultaneously, while each surgeon can observe the procedure in real-time.
Although it is challenging to incorporate large parallel computing arrays into clinical equipment, the powerful computing capabilities of GPUs and Tesla make it possible to offer compact parallel computing modules.