AI software takes cardiac risk assessments to another level

He also advocates moving away from risk scores that summarize a person's risk of having a cardiac event in the next 10 years with a percentage. Instead, he thinks a staging system, similar to cancer or heart failure diagnoses, can be much more beneficial. Taylor said this would help patients take heart disease more seriously, as they often do when they are given a cancer diagnosis.

From probability to precision cardiac risk scoring

Traditional cardiovascular risk assessments rely on algorithms such as the ASCVD Risk Calculator, which provides percentage-based estimates of a patient’s 10-year risk for heart attack or stroke. While they do provide some value, Taylor says these models lack the accuracy and clarity needed to identify high-risk patients, especially women and younger individuals who are often underdiagnosed.

Telling a patient they have a 7.2% risk of having an adverse event in the next 10 years may not mean much to them, he explained. Telling them they have stage-four atherosclerosis, however, can get their attention.

Coronary CT angiography (CCTA) also provides detailed medical imaging of the patient's arteries to see what is going on under the hood, rather than guessing based on population-based risk scores. In fact, Taylor compared CT imaging or the coronaries to breast imaging of a lump to determine if it is cancer, benign or something to keep an eye on.

AI-enhanced imaging brings atherosclerosis into focus

Heartflow's Plaque Analysis, combined with detailed CCTA imaging, rapidly and reproducibly quantifies plaque volume and characterizes dangerous vulnerable, soft, low-attenuation plaques most likely to rupture and trigger heart attacks. Taylor said these features could be seen on CCTA for years, but it was a labor-intensive task, requiring hours of manual interpretation, calibering and quantification that was not easily reproducible. AI now automates the process in seconds.

From a business standpoint, Taylor explained that CCTA may be the gatekeeper to determine which patients need the newer, more expensive drugs on the market. Accurately identifying the right patients for these medications, he emphasized, is critical.

Going beyond calcium scores

While low-dose CT calcium scoring has long been a staple in cardiovascular screening for several years, Taylor believes its role is evolving. 

“Calcium tells you the history of the vessel—it’s our best risk detection tool, but it’s not the full picture,” he explained. “The plaques that cause heart attacks are soft and non-calcified. We need to quantify total plaque and its high-risk features to truly stratify risk.”

AI is even making calcium scoring more accessible, with algorithms now able to automatically generate scores from routine CT scans done for other reasons, like lung cancer screening.

The case for widespread CCTA screenings

Despite the promise of this technology, clinical guidelines have yet to endorse widespread CCTA screening for asymptomatic patients. Currently, AI plaque analysis is primarily used for patients with symptoms or known risk factors. However, Taylor believes this will change.

“Ongoing trials will help answer the question of whether CCTA screening is superior to current risk models or calcium scoring,” he said. “But we already know the status quo isn’t working. Many patients who suffer cardiac events were never flagged as high-risk, and those who could benefit from intensive therapy often don’t receive it.”

He added that the continued underdiagnosis in women and younger adults is a persistent and unacceptable problem. 

“We’ve had PCSK9 inhibitors for a decade, and yet only 4% of eligible patients are getting them. That’s a travesty,” he explained.