IT News Archives | Site Search | Advertise on IT Direction | Contact | Home

IBM helps doctors detect life-threatening problems

May 18, 2009

Add to del.icio.us     Digg this

IBM has developed new technology that helps the medical profession in performing perpetual analytics like continually detecting subtle physiological changes in premature infants at the University of Ontario Institute of Technology (UOIT).

IBM's new solution, dubbed "System S" will assist UOIT's doctors detect life-threatening conditions and make better decisions by continually observing and collecting physiological data streams like respiration, heart rate and body temperature.

Nagui Halim, chief scientist for stream computing at IBM says "our solution to problem solving aims to take a bite-sized approach to ingesting data at different stages as it becomes available and continually refining the outcome, instead of the traditional monolithic stack."

Carolyn McGregor, a professor with UOIT says "the goal isn’t to replace the traditional manual paper-based approach employed by many medical practitioners. Rather, the technology will supplement those methods because that extra layer really represents a change in the fundamental way that critical health care is administered.”

McGregor explained that a progressive approach to analytics would entail first performing data encryption on the voice, then voice to text conversion, followed by keyword extraction.

“We allow programmers to specify how data is processed to this pipeline organization,” she said.

“The idea is to allow IBM's application developers to think about processing information in this progressive manner,” said Halim.

IBM made available at no charge System S' trial code to help get customers acquainted with the novel idea of stream computing. The flow model solves problems that are not necessarily those that enterprises have been able to solve in a general way, said Halim, and “we feel it’s important to introduce customers gradually to this.”

Additionally, TD Waterhouse Securities is using System S with the goal of making faster financial trading decisions by ingesting trading data at the rate of more than five million bits per microsecond. IBM said the equivalent would be a stock trader reading the entire works of Shakespeare 10 times in less than one second and then identifying and executing a stock trade faster by about 1,500 times.

In a typical flow model, there potentially could be thousands of concurrent streams where results are joined at the end or some other point, explained Halim, creating what can be a “fairly elaborate graph.”

For this reason, he said IBM created a means of expressing this flow model, a new development environment, “and the system takes care of all the difficult issues of scheduling, failover, congestion control, any basic things that a runtime has to do.”

Chip-maker AMD also has its own ATI stream technology, originally developed by ATI Technologies that enables graphics CPU's to work like an application accelerator for a variety of use cases, said Patricia Harrell, director of stream computing for AMD.

“What stream computing does that’s different is it gives us the opportunity to use that compute capacity for things other than traditional graphics,” said Harrell.

IBM also announced the opening of its new European Stream Computing Center in Dublin, Ireland, that will serve as a hub of research, customer support, and advanced testing for stream computing applications for European businesses.

Other applications of the technology at the lab will focus on water management, specifically looking at the health of local water ecosystems that intersect with human populations.

“System S can be connected to sensors that would detect water quality, temperature, salinity or fish that could be in the water in order to get a better sense of what’s going on,” said Halim.

For example, AMD’s stream computing technology is being used to accelerate algorithms to reconstruct CAT-Scan data, said Harrell.

“When you take a CAT-Scan of a patient, there is lots of post-processing you need to do to turn that data into an image that a radiologist can look at,” Harrell added.

She said that previously "there have been a number of different architectural approaches taken to accelerate applications, but some of them in the past have been very special, complicated technologies, some of them have been hard to program.”

Harrell thinks that stream computing is certainly gaining traction fast, especially considering the number of game-changing things happening, like Open-TL, an industry standard programming model for stream computing.

“Now mainstream developers can develop code that take advantage of those capabilities, and it will run on any server platform where the vendor also supports that standard,” said Harrell. “That needed to happen to make this go mainstream.”

Until now, only point solutions with limited capability have existed in the past. Harrell believes that stream computing will break us out of that pattern to enable more sophisticated applications in a variety of areas like location-based services, security and transportation.

Add to del.icio.us     Digg this

Source: IBM.