New research from a Bayesian network of network experts shows how Juniper Network can save your life.
The researchers behind the research have discovered a way to prevent a catastrophic cardiac event by predicting which people would be most likely to die at a particular point in time.
The network was created by an engineering team from the University of Texas, Austin, and uses a combination of a “skeleton” algorithm that combines hundreds of thousands of simulated observations to determine the likelihood of a heart attack.
“It’s a combination that we call a Bayes factor, and it’s a way of determining the probability that a particular person will die,” said professor and lead author of the research, professor of electrical engineering Dr. Paul Z. Davenport, in an interview with RTE News.
“In our experiments, we used the model to predict the chance of a person dying in the event of a cardiac event and then we used that information to treat patients.
We were able to save a life.”
Davenport explained how the network works.
“We start by generating a set of data from a random set of people,” he said.
“These are people who have had cardiac events, or people who are healthy, and we then compare those two sets of data with a set from the same random set to see how much of each one of the data is the same.
And that is where the algorithm comes in.”
To test the effectiveness of the network, the researchers compared the outcomes of different patients treated with a pacemaker to see if the algorithm could predict the likelihood that one of them would die.
The algorithm’s predictions were highly accurate, which was surprising because it was often the case that the patients in the trial were the least likely to live.
“One of the key things that we noticed in our experiments was that the more the model was able to predict, the more accurate it was,” said Dr. Darnport.
“For example, we showed that when a patient with a history of heart disease was in the early stages of a new diagnosis, the algorithm was able at that point to predict that they would be at the lowest risk of dying.
In contrast, the prediction accuracy for people with no history of cardiac disease was extremely low.”
After a patient died, the algorithms’ prediction accuracy was reduced.
This is because the algorithms are able to work backwards to determine which patients are more likely to survive.
When the algorithm is able to detect the risk of death at any point in the process, the risk reduction of patients is significant.
“The system is able also to predict whether the patient is more likely or less likely to suffer cardiac events.
And so the algorithm can then take the data and build a model that gives the patient the best chance of survival,” Dr. Nader said.
The results of the study, which were published in the Journal of Emergency Medicine, show that the algorithm does a better job than chance when predicting a cardiac death.
“When we compared the predictions of this algorithm with those of a patient without a history, the system was able predict that the patient was more likely, and the patient who was older was also more likely than the patient without cardiac disease,” Dr Z.E. said.
To prevent a cardiac episode, the research team had to make predictions that were both accurate and specific.
“So we had to understand that the data that we were getting from the algorithm at any given point in that algorithm was highly specific and the system could make the best predictions,” Dr Davenpont said.
As a result, the team used that data to identify the people with the most severe cardiac events at that particular point.
When that information was combined with the patient’s age and heart condition, the results were extremely accurate.
In the study’s second phase, the scientists used the algorithm to predict a heart event for patients with no prior history of a stroke or heart attack, and also for patients who had a history and were under a cardiac stress test.
When they were asked to make an accurate prediction, the patients who were younger had a lower chance of surviving.
“I think what’s really exciting about this is that the algorithms can predict the outcomes in such a way that the individual is more or less protected,” said Professor Daven.
“And so we’re hoping that this will be the kind of tool that doctors can use to get the most out of the technology in the field.”