Cardiologist Andrew Ambrosy Sees Promise in A.I., but Tells Patients to Proceed With Caution

Andrew P. Ambrosy, MD, MPH shares advice regarding A.I. health tools

SAN FRANCISCO, July 9, 2026 /PRNewswire/ -- The recent arrival of the first wave of A.I.-powered health tech has the medical community genuinely excited about how the adoption of artificial intelligence can help detect many health-related issues, including heart problems, before they turn into emergencies. One of the most promising examples of "health AI" uses deep learning—a type of algorithm that learns to recognize patterns—to find coronary artery calcium, or CAC, on scans you may already have had for another reason.

CAC is simply calcium that builds up in the arteries feeding your heart, and the more of it you have, the higher your risk of a heart attack or other heart trouble down the road. Heart patients, once they are discharged from the hospital or are in between office visits with their doctor, wouldn't normally have access to scans that would detect such CAC buildup.

With health AI, now they have the ability to constantly monitor this condition.

It's no small advance. Heart disease remains the leading cause of death worldwide, and much of it is preventable when we catch the warning signs early. Because these AI tools can read scans that were ordered for other reasons — say, a chest CT to check the lungs — they offer a chance to flag heart risk in people who never would have been tested otherwise.

Here is how it works. When you get a CT scan of your chest, the calcium in your heart's arteries shows up as bright spots. Traditionally, a specialist would have to measure that calcium by hand, which takes time and rarely happens unless a doctor specifically orders it.

Deep-learning algorithms can now do this measurement automatically — in seconds — turning scans that were already sitting in your medical record into an early-warning signal for heart disease. Several research groups and companies are developing and testing these tools, and the early results are encouraging.

The appeal is straightforward: much of the information we need to gauge a person's heart risk is already hidden in scans they have had for other reasons. AI gives us a practical way to surface it.

Used well, this technology could help identify people at risk years before they ever have symptoms—and prompt cholesterol, blood pressure, and lifestyle steps that prevent heart attacks in the first place.

That said, this is still a young technology, and it has real limits worth understanding. A CAC result from an AI tool is a helpful clue, not the whole story. Here are a few things to keep in mind:

• A "clean" scan isn't a clean bill of health. When AI checks your heart using a scan you got for another reason, it mostly spots hardened, calcified plaque. Softer, newer plaque—which causes many heart attacks—can be invisible to a calcium scan. So a low or zero score is reassuring, but it is not a green light to ignore chest pain, breathlessness, or a strong family history.
  
  
• The computer may not have learned from people like you. AI is only as smart as the patients it studied. If it was trained mostly on one type of person, it can be less accurate for women, younger adults, or certain racial and ethnic groups—and the result on the screen won't warn you. A confident-looking number can be shakier than it appears.
  
  
• A scary result can hurt as much as a missed one. Getting an alarming heart number with no doctor there to explain it can mean sleepless nights and a scramble for tests you may not need. AI is great at spitting out numbers; it's useless at the conversation that tells you what they actually mean.
  
  
• Nobody's cracked how to break the news. AI can now flag heart risk from a scan you already had—but finding the problem is the easy part. Telling you in a way that gets you to act without scaring you and giving your doctor a clear way to bring it up is something the medical world is still figuring out. The same result can land as a helpful nudge or a panic attack depending entirely on how it's worded. Right now, that's often left to chance.

Don't get me wrong. I find this work genuinely exciting. CAC is one of the strongest predictors we have of future heart attacks, yet today the calcium visible on millions of CT scans goes unmeasured, simply because no one was looking for it because the patient has either been discharged from a hospital or is between medical checkups.

Letting AI read those scans while the heart patient is at home, could bring this powerful information to far more people — at essentially no extra cost or radiation, since the scan was already done.

AI can erase missed opportunities and help close the gap and steer more people toward preventive care.

So, my advice is simple: welcome these results as a useful starting point but talk them over with your doctor rather than acting on a number alone. AI can open the conversation; it shouldn't have the last word.

That is one lesson you can certainly take to heart.

ABOUT DR. AMBROSY

Andrew P. Ambrosy, MD, is a physician-scientist at the Kaiser Permanente Northern California Division of Research, with clinical and research expertise in cardiovascular, renal, and metabolic conditions. Dr. Ambrosy's work focuses on generating real-world evidence through health services research, delivery science, and both traditional and pragmatic randomized controlled trials. 

SOURCE Andrew P. Ambrosy, MD, MPH