Research Highlights / Full Story

Due largely to the popularity of computer games, graphics processing units (GPUs) have become low cost, flexible in application, and able to run workloads well beyond 3-D graphics rendering. This has greatly enhanced various types of processing and visualization, including medical imaging. In the context of medical image segmentation, GPUs allow for the generation of models from volumetric scans, and greater localization of cancerous nodules during treatment, dramatically reducing radiation exposure to healthy tissue.

An RIT research team is working with Carestream Health and Roswell Park Cancer Institute on the improved use of GPUs in medical imaging for the advancement of image segmentation technology used in the process.

The team is currently exploring the GPU implementation of state-of-the-art medical imaging algorithms to speed up the execution of image segmentation. Initial analysis indicates that average segmentation time was reduced to less than one second, which corresponds to an increase in speed of over 50 times.

"Image segmentation is used to partition a digital image or volume into meaningful segments, making the representation easier to analyze and interpret," notes Andreas Savakis, professor and chair of the department of computer engineering at RIT. "When localizing cancerous nodes, the segmentation needs to be near instantaneous to accurately treat tumors during radiation therapy and differentiate between cancerous and healthy tissue."

The collaboration is facilitated by RIT's Corporate R&D program. Companies can receive applied research assistance that addresses specific design and development issues while retaining intellectual property rights for work developed through the project. This unique effort is designed to enhance university-industry collaboration and accelerate product development and commercialization.

"Our current collaboration with RIT will assist us in enhancing our medical imaging product line, while also advancing overall research in the field," says Dr. Jay Schildkraut, senior scientist at Carestream Health and a member of the research team.