If you go to a neurological ICU during a consultant’s morning rounds, you might observe doctors performing laborious tests to determine each patient’s level of consciousness. These tests are the only way to accurately predict a patient’s prognosis or find important warning signs that a patient’s health is getting worse. However, because each test can take up to an hour to finish, they put a lot of stress on clinical teams.
Researchers at Stevens Institute of Technology have now created an algorithm that can precisely track patients’ levels of consciousness using basic physiological markers that are already regularly monitored in medical facilities. Although it is still in its early stages, the team’s work, which was published in the Sept. 15 issue of Neurocritical Care, promises to greatly reduce the burden on medical staff. It could also give important new information that will help doctors make decisions and make it possible to come up with new treatments.
Samantha Kleinberg, an associate professor in the department of Computer Science at Stevens, explained that consciousness is more like a dimmer switch, with varying degrees of consciousness throughout the day. “You only get one data point if you check on patients once a day. You could continuously track consciousness using our algorithm, which would give you a much clearer picture. ”
With the help of Jan Claassen, director of Critical Care Neurology at Columbia University, Kleinberg and her Ph.D. student Louis A. Gomez developed an algorithm that forecasts the outcomes of a clinician’s evaluation of a patient’s level of consciousness. The algorithm used data from a variety of ICU sensors, ranging from basic heart rate monitors to sophisticated devices that measure brain temperature. The results were astounding: using only the most basic physiological data, the algorithm proved to be just as accurate as tests conducted using expensive imaging technology, such as fMRI machines, and only slightly less accurate than tests conducted using trained clinical examiners.
The fact that this tool could potentially be used in almost any hospital setting, rather than just neurological ICUs where they have more advanced technology, is crucial, according to Kleinberg. She said that the algorithm could be added to existing bedside patient monitoring systems as a simple software module. This would make it fairly cheap and easy to use on a large scale.
Continuous monitoring could lead to more research and, in the long run, better outcomes for patients. It could also give doctors better clinical information and help patients’ families understand how their loved ones are doing.
Because there is so little available data, studying consciousness is extremely challenging, according to Kleinberg. One day, doctors may be able to help these patients much more effectively if they can keep track of how their consciousness changes over time.
More work must be done before the team’s algorithm is implemented in clinical settings. The team’s algorithm was developed using data that was gathered just before a clinician’s assessment, and more work will be required to demonstrate that it can accurately track consciousness 24 hours a day. More information will also be needed to train the algorithm for use in other clinical settings, like pediatric ICUs.
By comparing various types of physiological data and examining how they track or lag one another over time, Kleinberg also hopes to increase the algorithm’s accuracy. Since some of these relationships have been shown to be related to consciousness, it may be possible to verify the algorithm’s ratings of consciousness when clinical assessments by humans are not available.
For the time being, however, the Stevens’ team is ecstatic to have discovered a straightforward, widely applicable model for automatically determining patient consciousness in clinical settings. A high-risk, high-reward project, according to Kleinberg. It was very exciting to find out that we could use these signals to group patients’ levels of consciousness.
