The Technion Introduces Smooth Surfaced 3D Medical Object Imaging

in Technion by David Miron — May 16, 2011 at 10:58 am | 0 comments

Share

If you’ve ever been pregnant, suffered from injuries, or experienced abdominal pains, then you’ve probably undergone some kind of medical imaging procedure such as an Ultrasound, MRI, or CT. Medical teams all over the world consider these methods as their first, and foremost tools for precision diagnostics. The medical Imaging world is now facing the challenges of applying new technologies for providing superior imaging services which will rewrite the standards of the trade. This is especially so with 3D multi-labeled objects.

Creating 3-D surface object in medical imaging

Medical imaging applications use 2D image slices which are based on contour information. An image of a region is represented by many such two dimensional input slices (see image below), that are then be used to render a three dimensional reproduction of the object. This process is far from straight forward, and complicated algorithms are required to create the final product. To come up with a smooth surface 3d object, a reconstruction algorithm needs to be applied to a set of planar cross section slices. In reality, the two dimensional input slices are often filled with noisy information of unclassified contour regions, and are considered low quality for the purposes of 3D object rendering.

Simplifying the 3D object rendering process

Technion researchers Amit Bermano, Craig Gotsman, and Amir Vaxman have developed an advanced reconstruction algorithm extension which can deal with the ‘noise’ orginitaing from low quality input data. The algorithm treats input containing unknown regions, in and around, the object, and then enables the rendering of smooth surfaced 3D images of the target object. This unique reconstruction algorithm extension can handle the cross sections data even when slice regions are classified as being inside the object, outside the object, or unknown. The end-result is smooth interpolating surface corresponding to a set of segmented cross-sections even if they contain unknown regions. This process is achieved in a minimal amount of time.

Interactive medical applications use online reconstruction

Asides from the ability to reconstruct 3D objects from arbitrary data with minimum time, and complexity, this unique algorithm is highly parallelizable. As a result, the ability to generate an efficient online version, performed on a Graphics Processing Unit (GPU) suitable for interactive medical applications, is now feasible.

Those interested in this technology for commercial purposes are welcome to Join the Technion’s EIR program.
The EiR program introduces entrepreneurs to the Technion’s research environment, helps them, identify promising technologies, and assists in launching Start-up Companies built around these technologies.

For more information contact Inbal Lev, Director of Business Development at the Technion. inball@dp.technion.ac.il

This post was written by Michal Gaziel-Yablowitz of the Technion

Tags: featured Technion