by Gus Iversen
, Editor in Chief | November 23, 2015
Q&A with Dr. Dorin Comaniciu Vice President of Siemens’ Medical Imaging Technologies
The images you’re looking at on the following pages may look like something out of a video game or science fiction movie, but there’s nothing fictional about them. These are highly detailed renderings of actual living individuals — and they were produced using Siemens’ cinematic rendering technology.
To see a portfolio of dramatic cinematic rendering images, courtesy of Siemens Healthcare, please click here
These stunning depictions represent the cutting edge of data visualization, and HealthCare Business News spoke with Dr. Dorin Comaniciu, vice president of Siemens’ Medical Imaging Technologies in Princeton, New Jersey, about how this technology came to be — and what role it may play in the future of medical imaging.
HCB News: What is cinematic rendering technology?
Dr. Dorin Comaniciu: Today medical imaging scanners, like CT and MR, deliver very detailed data representations of the interior of the human body. Cinematic rendering is an innovative algorithm that addresses the visualization of such data in order to provide superb photorealistic images. The inspiration for our scientists was the film animation industry: design a new rendering technology that helps to create visualizations that are increasingly realistic. We believe such technology opens new possibilities for medical education, early disease detection, planning and guiding of surgical procedures, and ultimately facilitates better communication between referring physicians and patients.
HCB News: How was it developed?
DC: Our scientists asked the question: how can we advance the visualization of medical data? The images provided today by MR and CT have incredible resolution and so many details that are useful to the physician. However, sometimes they are difficult to visualize. To understand the complexity of this data, imagine a large-volume CT or MR of the whole body. You can think of a cube made up of several hundred million voxels (i.e., elements of the 3D Cartesian grid), each voxel characterizing specific information about the structure within the human body, such as density or tissue type. How do you visualize such a large amount of voxels? How do you make sure you properly see the rich content present in this data? Assuming light in the surrounding environment can reach any location from a myriad different directions, the main idea is to model the light propagation — the interaction of the light rays with the data — and to compute and display the resulting properties of the rays that reach the human observer.The integral equation that models the underlying geometric optics is quite complex.However, it can be solved iteratively through a process known as Monte Carlo integration. This is essentially what we implement through cinematic rendering: To reveal the volumetric information of interest, we iteratively propagate billions of light rays through the original volume, and due to parallelization and optimization algorithms such as importance sampling, this process runs in, close to, real time. In this way, by modeling the light scattering and the bouncing of rays, one can accurately represent the content of volumetric data. We can observe soft shadows, depth of field, subsurface scattering, caustics, motion blur… everything that makes the images very realistic.