The patient says she has trouble breathing, even when she’s resting. She has a stubborn cough, her ankles are swollen, and she’s tired all the time. Recognizing the constellation of symptoms describing heart failure, the patient’s doctor orders an echocardiogram. The test reveals there’s something wrong with the patient’s heart muscle, raising the need for a more precise diagnosis. Is the patient suffering from cardiomyopathy or pericarditis?
Distinguishing between these two diseases is tricky. To make the right call, heart doctors rely on their experience and the wealth of data from echocardiograms. But even with the combined power of training and technology, results are mixed. A skilled and experienced cardiologist can make an accurate diagnosis three out of four times. For other physicians, the accuracy rate is closer to one in two. Not much better than a coin toss.
Fortunately, new technology is available to help doctors make the call.
“The great benefit of AI for healthcare is the democratization of patient experience.”
In a recent experiment, researchers using AI spotted the difference between pericarditis and cardiomyopathy nine out of 10 times.
The successful test holds the promise of giving physicians a valuable diagnostic tool that keeps patients healthier while also reducing costs for providers. Here’s how the experiment worked:
The team loaded data from 50 patients with constrictive pericarditis and 44 patients with cardiomyopathy into the Intel® Saffron™ AI. That’s a staggering quantity of data: 90 measurements taken from six locations in the heart and 20 times in a single heartbeat for a total of 10,000 attributes per patient per heartbeat1.
“Before, for these types of diagnoses, you had to know what you were looking for; focus in on a few attributes—and in the process, human bias was introduced. The revolution here is being able to deliver higher accuracy based on all 10,000 attributes per heartbeat. It’s a no-compromise scenario, which results in the best outcome,” said Gayle Sheppard, Intel’s vice president and general manager for artificial intelligence.
“Before, for these types of diagnoses, you had to know what you were looking for; focus in on a few attributes—and in the process, human bias was introduced. The revolution here is being able to deliver higher accuracy based on all 10,000 attributes per heartbeat. It’s a no-compromise scenario, which results in the best outcome.”
The researchers used associative memory AI, which mimics how the human brain works to surface similarity patterns. These solutions are designed to help people in highly complex tasks, and require less data to learn than traditional solutions; they can learn and adapt in real time without predetermined rules and parameters, and can yield insights faster.
Not only did AI achieve an accuracy rate of more than 90 percent, it did so with a shorter learning curve than other machine learning approaches. It reached its accuracy rate after ingesting data from only one-third of the patients.
Artificial intelligence is transforming every industry, but the role it will play in healthcare is profound, said Sheppard.
“The great benefit of AI for healthcare is the democratization of knowledge. AI can provide more wisdom of experience to physicians and patients about complex diseases, about treatments, and the outcomes that are achieved,” Sheppard said. “What if you could open up all the knowledge of the world, all the experience of physicians around the world, and make that available to any physician anywhere in the world?”