Using data stream analysis in brain research at UCLA’s School of Medicine
I recently spoke with Dr. Paul Vespa, a professor of neurosurgery and neurology as well as director of neurocritical care at the University of California’s Los Angeles (UCLA) campus. The kind of work that Dr. Vespa and other colleagues do involves brain research (they collect vital signs information using a variety of sensors and track the dynamics of brain functions). Further, these doctors provide care for patients with critical brain injuries.
I met with Dr. Vespa was to learn about how his organization is making use of new IBM data streaming technologies to gather and analyze data in order to better understand what is going on with a patient’s brain—here’s what I learned:
- Doctors at UCLA have recently started using data streaming technology in order to make more informed decisions about brain functions and abnormalities.
- UCLA is using IBM Watson Foundations, a collection of data organization and management facilities combined with analytic tools, to collect and analyze information pertaining to patient brain functions. More specifically, IBM Watson Foundations is being used to provide physicians with decision support for brain analysis.
- Using IBM Watson Foundations physicians are able to gather data from sensors to analyze brain functions in real time. This real time element is made possible by capitalizing on IBM’s ability to analyze streaming data—data that doesn’t necessarily go to disk where it can be subsequently analyzed at a later point, but instead is analyzed by the computer system as it is fed into that system.
- As a result of the use of this technology, patient care can be substantially improved, and doctors have more time to serve more patients.
A big improvement
UCLA has been conducting medical research for decades, so the natural question to ask at this point is: “How does this Watson Foundations approach differ from the approach that was being used previously?” Dr. Vespa’s response to this question indicated that before UCLA was able to acquire Watson Foundations streaming services, the previous approach was more manual in nature, and somewhat inflexible. Using the previous approach, data from sensors would be collected, but it consisted mostly of just numbers (raw data presented in standard, static data tables). “These tables did not give us the information on what these numbers mean” said Vespa. Third party analytics tools, which Vespa described as one-size-fits-all were then used to analyze this data.
In contrast, the new IBM Watson Foundations approach helps make sense of the data that is collected in real time so that brain variables can be analyzed and displayed second-by-second. Various forms of data from a variety of sensory sources are fed via streams to a computer system that can analyze this data in parallel, helping to quickly paint a broader picture of what functions are taking place in a brain. Using this real time approach, information is parsed so only the important data is being analyzed; it can be time aligned to see what is taking place in the brain as a whole at any given moment; and from this analysis a multi-dimensional one-page visualization of brain activity can be produced. As a result, doctors gain a deeper understanding of brain activities and interdependencies and, as a result, are better able to diagnose and treat brain injuries.
Issues and challenges
UCLA is one of eight educational institutions using this streaming approach in medical analysis, and, as is usually the case when new technologies are deployed, there are some initial challenges that must be overcome.
The first issue that arose when setting up UCLA’s streaming solution related to sensor feedback. Various sensors provided by various vendors present data in various ways (there are few common standards for presenting sensory data). This data must be interfaced (aligned with the database) such that it can be read. So UCLA’s first challenge was to create an interpretation interface to overcome data presentation challenges. The good news here is that the eight educational institutions mentioned above are cooperating with each other and sharing information about interfaces. So, interface challenges are being minimized and an ecosystem is evolving to help address these types of issues.
The second issue is common to analytics across the entire computing spectrum: where are the resources to be found that can build analytics programs and structure queries? In Dr. Vespa’s case, the types of analyst skills needed to build the analytics environments necessary to develop a picture of brain activities based on a wide number of parameters can be found in engineers who understand computational mathematics. Being a university, UCLA has access to these skills, but they are not easy to come by. Still, as was the case with the interface problem mentioned above, an ecosystem is developing and programs are being shared, so this issue is slowly being ameliorated.
One final issue is ease-of-use. According to Vespa, “physicians don’t want to spend the time learning how to use a clunky tool.” So if the interface to the back-end streaming data is complex, physicians might just run with their own interpretations of the data already have at their disposal. Ease-of-use improvements are constantly being made to the UCLA Watson Foundations platform in order to encourage more widespread use.
When queried about the future of IBM Watson Foundations at UCLA, Dr. Vespa abounded with enthusiasm: “right now, we have a real problem with alarm fatigue: false alarms that require swift medical attention. So we’re working with AlarmX, a product that can analyze streams and can detect cardiac arrhythmia. Using this combination of AlarmX and IBM Watson Foundations [for streaming support] we can get a better, more accurate sense of critical alerts. I expect more and more products like AlarmX to come to market that will help streamline patient care and free us up to help more and more people”.
As for other futures, Vespa expects more standardized information to be made available for streams analysis; the more feeds of data, the better the diagnosis and treatment can be expected. “What we’re doing is building out our informatics base in order to take better care of patients” he stated. He also expects that analytics technologies will become highly instrumental in medical forecasting and in pre-emptive medicine.
Doctors need access to disparate information from a number of sources in order to best diagnose and treat injuries and ailments. Using real-time streaming, doctors can now see the logical and interrelated progression of events within the brain, and make diagnoses more accurately. I see IBM as the leader in data streaming technology, and with products like Watson Foundations combined with ecosystem analytics solutions from third parties, I see IBM positioned to pull away from its competitors by a wide margin.
I’ve seen nothing quite like Watson Foundations coming from any other systems or database vendor in the industry, so if you have any data to the contrary, please comment on this blog. Meanwhile, explore my other posts on IBM Watson Foundations here on the Hub:
- An analyst's examination of IBM Watson Foundations
- nViso: Using IBM Watson Foundations to read emotion