'Stippling' may allow real-time 3D imaging

December 9, 2002

The ancient art of stippling -- the placing of small dots on a surface to create a picture -- may find its way into digital imaging.Purdue University computer scientists have devised computer imaging software that uses stippling to produce pictures of

The ancient art of stippling -- the placing of small dots on a surface to create a picture -- may find its way into digital imaging.

Purdue University computer scientists have devised computer imaging software that uses stippling to produce pictures of internal organs. The researchers claim the technique is 10 times faster than some conventional methods and could provide a tool for medical professionals to preview CT or MR images in real-time while a patient is being examined.

In stippling, also known as pointillism, the artist creates numerous dots to produce gradations of light and shade, forming an image. The most famous example of pointillist art is A Sunday Afternoon on the Island of La Grande Jatte by the 19th-century French painter Georges Seurat.

Because dots are the most simple visual element in a picture, they also are ideal for computer visualizations, said David Ebert, Ph.D., an associate professor in Purdue's School of Electrical and Computer Engineering.

"We take volumetric data sets from a variety of sources (converting them from DICOM to an internal format), then allow the user to load the data sets into the system and interactively view them," Ebert said.

The control panel can be used to emphasize different portions of the data via rotating and zooming with a mouse. Users can highlight organ boundaries or see more of the volumetric interior of the data. They can also make images that look like illustrations through the use of silhouette enhancement and oriented lighting effects, he said.

"More conventional imaging methods of, say, a CT scan of a person's head, require slower processing techniques. We can have a CT rendering of a person's internal organs in real-time, where the organs are represented as a series of small points," Ebert said.

Ebert has received positive feedback from radiologists at the University of Colorado at Denver who have seen the system. Other radiologists reacted with skepticism, viewing the Purdue method as just another volume rendering technique.

"My initial reaction is that this is probably not going to be very useful," said Dr. Bradley Erickson of diagnostic radiology at the Mayo Clinic in Rochester, MN.

One of the big problems in 3D for medicine is segmentation and classification -- defining which pixels belong to which organs. Once that is done, arbitrarily fast rendering (at reduced quality) can be achieved by reducing the detail or number of surfaces. The Purdue work appears to be a rendering method and does not solve this problem, Erickson said.

Once segmentation is done, one can generate points, or triangle meshes, or polygons to be rendered, he said. Since most "fast" consumer-grade video cards do polygon rendering in hardware, quite good performance can be obtained with polygons at a low price.

"The images provided do not appear to have the quality necessary to do anything beyond very gross navigation -- something that can also be accomplished with large polygons at only a very small fraction of time spent working with 3D images," Erickson said.

Ebert said he is interested in working with radiologists and physicians to develop the system into a more useful tool and to determine its utility in medical applications.