Researchers used a smart computer program, similar to those used for chatting online, to figure out the best ways to prevent strokes. They filled the program with data from lots of patients, all kept anonymous, from places like workplaces, health plans, and hospitals.
They trained the program to understand different health issues, like the risk of having a stroke, by giving it lots of general information first. Then, they fine-tuned it to focus on stroke prevention and tested it against other similar programs. Their program did better and even matched suggestions from real medical trials.
The researchers don’t plan to replace regular medical research, but they think their program could help speed up trials and make treatments more personalized, which could save time and money.
Their program could be like a fast track for finding which drugs might work for a disease. It could help doctors run small trials with just a few drugs to see which ones are best for patients.
The researchers named their program CURE, short for Causal Treatment Effect Estimation. They trained it using a huge amount of real patient data from 2012 to 2017, including medical codes and medication information.
They made their program even better by adding extra medical knowledge to it. This helped the program understand patients’ data even more.
In the future, doctors might use this kind of program to help them decide the best treatment for a patient by looking at their medical history. But first, it needs approval from the FDA, the agency that makes sure medicines and medical tools are safe.
The research was funded by the National Institutes of Health.
Source: Azorobotics
