Machine Learning Bypasses Need for Some Invasive Procedures
By Jane Ehrhardt
A new noninvasive diagnostic technology can quickly assess the likelihood of coronary artery disease (CAD) and pulmonary hypertension (PH) at the time of evaluation. Arriving first in Alabama at Cullman Internal Medicine, a 23-physician practice, the CorVista system utilizes machine learning to analyze a patient's cardiac signals, bypassing the need for invasive procedures to identify potential several cardiovascular issues at the point of care.
Cullman Internal Medicine has referred patients to the cardiologists at the Cullman Regional Medical Group Cardiology Clinic (CRMGCC). “We have now employed CorVista technology in our cardiology office, and also in the hospital setting, to help us streamline diagnosis,” says Tracy Neal, MD, a cardiologist with CRMGCC.
In the two months since the clinic started using CorVista, they have already seen the device’s accuracy save a life. A 72-year-old patient had been evaluated for atypical chest pain and underwent an evaluation by a cardiologist and a nuclear stress test. The stress test came back normal. However, the patient’s symptoms continued.
“So we ran the CorVista test on this man and it was abnormal, indicating the possibility that we had missed something,” Neal says. “We performed a heart catheterization and found a 99 percent blockage in the main artery down the front of the heart. That's called a Widow Maker blockage. CorVista was able to pick up a heart problem that we had previously missed and saved the man’s life.”
Currently, the CorVista System holds FDA approval to detect the likelihood of coronary artery disease (CAD) and pulmonary hypertension (PH). The company is currently seeking approval for identifying pulmonary capillary wedge pressure.
The system works similar to an EKG. “Whenever the heart beats, it gives off millions of electrical signals, but it also emits millions of other physiologic signals such as impedances, which are resistances within the way electricity flows,” Neal says. “And the heart has a normal movement and a normal vibration.”
The device’s seven sensors attach to the patient to collect specific electrical and physiologic data signals. The information arrives at the handheld device, about the size of an iPad, and the physician sends it along to CorVista’s cloud system where advanced machine learning interprets the results in comparison to data from 11,000 cardiac patients who had undergone catheterization. Then it sends a report back to the handheld device. The whole process takes about 15 minutes.
“It’s a binary result,” Neal says. “It’s either positive for the condition or negative for the condition. No test is completely accurate, but this test has an 88 percent sensitivity for determining whether someone has severe significant blockage or not. It also has a 99 percent negative predictive value for determining if someone does not have heart disease, which means that if the test says the heart is okay, there's a 99 percent chance that the heart is indeed okay.”
Practices do not buy the system or pay a subscription. Instead, the company provides the device, and they collect the insurance fee for reading the test, while the physician bills insurance for administering the test.
“It’s another good tool to help us streamline cardiac diagnosis, and it helps us look at a larger number of patients in a much more rapid fashion,” Neal says. “For primary care clinics, the system provides a valid evaluation on who needs to get in to see the cardiologist quicker. And it gives the practice a proven test to back up that request. We don't pay anything at all, so it was a no-brainer for us to incorporate it.”