IOD2009 – InfoSphere Streams

I’m still catching up on the events of last week’s IOD 2009 conference.  Those of us participating in the inaugural blogger program were treated to several “behind the scenes” meetings with executives, developers and IBM customers on a variety of topics, one of which was stream computing and in particular how it’s being used  by the University of Ontario Institute of Technology to help save lives.

I’ve blogged before about the initiative to use InfoSphere Streams to monitor premature infants in the neonatal intensive care unit, which has the potential to alert hospital staff of an impending infection in the preemies up to 24 hours earlier than it might have otherwise been noticed using current practices.  We were able to sit down with some of the researchers and developers behind InfoSphere Streams, as well as Dr. Carolyn McGregor, a UOIT associate professor and Canada Research Chair in Health Informatics and learn more about this effort first hand.  (This effort was also featured in a video during the Day 2 General Session – I’ll link to the video as soon as I can find it posted online.)

The line that really caught my attention was when Dr. McGregor described doctors as “walking backwards in time” when they walk into the neonatal intensive care unit – taking out pen and paper to record notes and make sense of monitors and diagnostics.  (It’s not uncommon to be using monitors and physical equipment that can date back 20 years; Imagine for a minute trying to do your job with the same computer you were using in 1989.)

Even with all of the equipment and monitors you can find in a typical intensive care unit (let alone a NICU), results and observations are often abstracted down to a single number recorded by hand on a grid every 30 or 60 minutes.  At best clinicians end up looking at just one stream of data, or focusing on just one patient.  This is in sharp contrast to research that has found that there can be significant improvements made in fighting infection by looking across multiple streams of data, multiple patients, and multiple diagnoses (infection often spreads through the ICU, so it’s especially important to be able to compare what is happening across patients).

This is exactly what the initiative undertakes to do – monitoring preemies on a continual basis and providing real-time alerts when analysis matches conditions shown by research to be an early indicator of infection setting in.  In contrast to a trend towards building dedicated monitoring boxes, InfoSphere Streams can allow for monitoring and analyzing multiple streams of information concurrently.

And this is just the beginning – imagine sharing monitoring data remotely with expert neonatalogists by regional hospitals who don’t have access to such expertise on site.  The implications on decisions related to courses of treatment or when best to transport an infant can have an enormous impact, as well.

A software-based approach to monitoring will also allow clinicians to introduce new rules into the environment, so they can discover and test new findings.

Stepping out of this particular example, the developers of the technology described what is different about the approach employed by InfoSphere Streams.  It enables streams of data to be split into pieces so that the software can process in parallel different steps of what is being monitored.  For example, a law enforcement agency might want to take data from a video camera, take a snapshot every few seconds, then do facial recognition to see if there’s a person on screen, and if so, match it against a list of people being sought out.  If it’s a match, use the GPS location of the camera to alert law enforcement in the area.

A system that can handle problems like this in this way  – progressive information processing and analysis along the way, bringing together streams of data from multiple sources at once – didn’t exist before; researchers started from the ground up, inventing everything (including the programming language) along the way.  Other projects that will use the analytics in stream computing include monitoring traffic or forecasting space weather.

On a personal note – thank you to everyone that came out at the end of a long IOD day to meet with us.  I know I (as did I’m sure my fellow bloggers) appreciated hearing more about this technology and initiatives first hand.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: