HeartFlow: Using AI to Fight Heart Disease
The older you get, the more you understand why elderly people spend so much time talking about their health problems. It usually starts around age 40, when you realize you have reached the age when you should get regular medical checkups. You get a blood test, and your doctor says your cholesterol is high.
So you switch to a "Mediterranean diet," buy a new running watch, but after a few weeks you slide back into your old habits and in the end simply arrive at the American solution to all problems: pills. Yet cardiovascular disease is no joke. It is the number one cause of death, accounting for one in three deaths, which is more than twice the number caused by the second most common cause, cancer.
Roughly every sixth healthcare dollar is spent on heart disease. Given the scale of the problem, it is no surprise that AI algorithms are emerging that could help us save lives.
Founded in 2007, HeartFlow is valued at $1.3 billion. Although the company's history actually began back in 1995 with the first image-based simulation of coronary blood flow. The funding was used to build an "AI-powered cloud enterprise suite" that will help with risk assessment, diagnosis, planning and treatment of cardiovascular disease.
The company's website says: "HeartFlow is a medtech company that has revolutionized precision heart care. No other company in the world does what we do, combining decades of human ingenuity and cutting-edge technology in a completely new way. We are passionate about transforming the diagnosis and treatment of coronary artery disease and continue to innovate, pushing boundaries and forging incredible partnerships."
Invasive and Non-Invasive Testing
Any competent doctor will likely notice early signs of heart disease risk during a routine checkup, such as high cholesterol or a family history of heart failure. A sufficiently high level of risk may require further non-invasive testing. In the US alone, 10 million such tests are performed.
Today's risk assessment methods are underdeveloped at best compared with the future vision of precision medicine. That is why HeartFlow goes beyond simply improving existing methods. The company aims both to improve the ability to correctly diagnose heart disease and to eliminate the need for unnecessary invasive procedures.
The invasive method is a procedure also called invasive coronary angiography. It requires inserting wires and catheters into an artery in the patient's wrist or leg, which are then advanced to the heart under X-ray guidance. An X-ray contrast dye is injected through the catheter and images of the coronary arteries are taken. This process is expensive and potentially problematic, since complications can occur. This is where HeartFlow comes in.
The HeartFlow Analysis
This is a cloud software platform that uses deep learning to create an accurate, patient-specific three-dimensional computer model of the patient's coronary arteries, based on analysis of non-invasive imaging data.
Computed tomography angiography (CTA) is a type of medical test that combines computed tomography with the injection of a special dye to obtain images of blood vessels and tissues in a part of your body. The dye is injected through an IV line starting in the arm and "lights up" the blood vessels and tissues being examined.
The key limitation of CTA as a standalone test is its inability to characterize the clinical significance of intermediate stenoses (artery narrowing of 30-90%). There is also an additional problem: 20-30% of false negative results (sometimes in patients with heart disease, they are not detected during CTA). HeartFlow addresses both problems.
After a CTA test is performed, the existing images are passed to HeartFlow's algorithms, which use computational fluid dynamics to simulate blood flow through 3D models of the arteries. HeartFlow's algorithms were trained on tens of thousands of CTA applications, including more than 15 million cardiac computed tomography images.
The end result is information that changed the patient's treatment in more than 60% of cases and made it possible to safely avoid invasive testing in more than 60% of cases. With HeartFlow, non-invasive testing can replace invasive testing and also deliver better results.
As of June 30, 2021, coronary blood flow analysis had been performed on more than 100,000 patients. The company's documentation presents results from many studies showing its platform's superiority over existing methods. The technology performs well: the likelihood of detecting heart disease with it is twice as high compared with other tests.
