A CiTIUS research will improve 3D Medical Imaging processing

Common medical procedures, as magnetic resonance or computer tomography, are only two particular examples of hospitals’ daily routine. The development of these and other diagnostic tools has allowed medical professionals to access to new and very valuable information for the diagnostic process. But in order to get the most of the data, results should be consulted and visualized in real time.

A CiTIUS researcher, Julián Lamas, has obtained his Phd degree after exposing his dissertation in the Center. In his work, J. Lamas has studied several strategies for the processing of 3D medical image, developing new techniques to maximize the performance of the algorithms (that is, the set of mathematical operations that are employed through the processing of medical image). Results have confirmed that it is possible to multiply the efficiency of some of the algorithms by a factor of more than 50. The advances will enable to compute high requirement computational processes in the hospital facilities using low cost hardware, specifically Graphic Processing Units (GPUs).

Faster and less invasive medicine

IMG_6731Processing 2D and 3D image is one of the pillars of the advance of modern medicine, but in order to meet the challenges of minimally invasive surgery or diagnostic imaging systems, new strategies to improve the performance are needed. Nowadays many systems and applications use costly algorithms that are not adapted to work in real time.

Julián Lamas’ Phd will contribute to the development of new systems for processing medical image in real time over common hardware (as the GPUs), reducing the cost of the systems and enabling their application to a greater number of patients.

Distributing computations and merging the results

This research has been focused in segmentation (which is used, for example, to identify organs and tissues, and in the detection of cancer nodules), as well as in visualization, with the objectives of presenting the different types of medical image in real time to the specialists. In the first case, the Phd has proved that is possible to improve the process by a factor of more than 50, while the visualization tasks have been improved between 5 and 12 times.

To do so, a new operator -called «divide and merge» has been proposed, in order to distribute the load between the different processors of the GPU and to combine the results produced by the different processors. Several strategies for visualizing compressed information has also been defined, as GPUs memory is not always enough to work with non-compressed medical image. Julian Lamas points out that «original algorithms have been heavily modified to obtain these results».

Research results will contribute to the diagnostic systems that are used by medical professionals. Some of the Phd solutions have been already implemented in the commercial software AMIRA, a leading software market solution.

Comments are closed.