Coronary heart disease (CHD) is the second most common cause of death in England and Wales and a condition that afflicts around 2.3 million people in the UK, but diagnosis of the disease is often complex and invasive.
Birmingham City Hospital has found an effective and safer alternative using deep learning software developed by US healthtech HeartFlow.
The Californian company has created a tool that uses deep learning algorithms to analyse cardiac CT scans and build personalised, three-dimensional models of the heart and connected blood vessels that illustrates the impact that blockages will have on blood flow. Physicians then assess the results to decide if enough blood is reaching the heart.
CHD diagnosis typically involves an invasive angiogram, which is used to check the blood vessels by inserting a catheter through an incision in the groin, wrist or arm, and up to the affected artery. A thin wire is then guided down the catheter to deliver a balloon to the affected section of artery, where it's then inflated to make the blood flow more freely. Dye is then injected to see how the blood flows. Sometimes the doctor will also insert a permanent stent, a wire-mesh tube that keeps the artery open. Common side effects include bleeding or bruising, but there is also a chance of allergic reaction, artery damage, heart attack, stroke or death.
HeartFlow substitutes the angiogram for a CT scan, a cheaper, safer and faster procedure that uses an X-ray machine to creates detailed images of the inside of the body. If the scan shows disease, the data is securely uploaded from the hospital's system to the cloud and sent to HeartFlow's headquarters in California, where the software analyses the image. It then creates a digital, color-coded, 3D model of the patient's blood flow that is sent back to the doctor who uses it to determine the best course of treatment.
Doctors can review HeartFlow's models on their tablets. Image creadit: HeartFlow
The software is proving its worth at Birmingham City Hospital, where Dr Derek Connolly, a consultant interventional cardiologist, had already sent three cardiac CT scans to HeartFlow the day we spoke to him.
"It effectively means that instead of bringing a patient in for an operation, they can be treated as an outpatient," he tells Computerworld UK.
A study published in the European Heart Journal found that using HeartFlow led physicians to reconsider and change their management plans for two-thirds of their patients. It determined that some could avoid a coronary stent or bypass operation, while others who would normally receive only medication were identified as needing surgical intervention.
"In a lot of the tests we do, the scan will be normal, but in a lot of the tests we do, the scan is clearly abnormal and the patient needs an angiogram and probably stents or bypass surgery," says Dr Connolly.
"What HeartFlow does is it allows us to assess the ones in the middle where we're not sure whether the narrowings we're seeing are flow limiting or not. And there it's really crucial because the eyeball isn't good. We've learned that historically from the invasive angiogram, and exactly the same is true of the HeartFlow exams. It helps us to decide who needs stents and who needs tablets and it's a great advance for us."
HeartFlow was developed by a mechanical engineer called Charles Taylor and a vascular surgery pioneer named Christopher Zarins. The duo spent 11 years commercialising their system.
As a moving organ, the heart was initially tricky to scan in real-time, but recent increases in computational power have overcome that barrier.
The AI also means that the results improve over time. Connolly has observed this at Birmingham City Hospital, where the system has now been used around 100 times.
"For those hundred scans, we've avoided people needing invasive angiograms on the borderline ones in about 70 percent of the patients," he says. "That means the 70 people out of 100 in our experience so far don't need an operation, and that is an enormous advance.
"But what it means as well is that time in the lab that we would have been using to do a diagnostic angiogram that didn't proceed to stents we can use for people who do need stents. We can find more people this way and treat the ones that need treatment and not treat the ones that don't need treatment."