What is hypertrophic cardiomyopathy, and why find it early?
Hypertrophic cardiomyopathy, or HCM, is a disease in which the heart muscle becomes abnormally thick. It is often inherited and can run in families. Many people live for years without knowing they have it, because it may cause few warning signs.
The authors note that HCM is uncommon but an important cause of sudden cardiac death. Doctors usually confirm it with an echocardiogram (an ultrasound of the heart) or a cardiac MRI, tests that take more time and cost more than an ECG. A tool that could flag HCM on a simple ECG might help send the right people for those tests.
How did the researchers build the AI?
The team used a convolutional neural network, a type of AI that is very good at recognizing shapes in images and signals. They trained and tuned it with 12-lead ECGs from 2,448 patients with confirmed HCM and 51,153 people without HCM, matched by age and sex. An ECG records the heart's electrical activity in about ten seconds using stickers placed on the skin.
The AI was then tested on a separate group it had never seen: 612 patients with HCM and 12,788 controls. The average age was about 55 in the HCM group and 57 in the control group.
What were the results?
Results are reported as area under the curve, a score from 0.5 (pure chance) to 1 (a perfect test). In the test group, the AI scored 0.96. It caught 87% of patients with HCM (its sensitivity) and correctly cleared 90% of people without the disease (its specificity).
Two findings stand out. First, the AI still scored 0.95 in patients whose ECG looked normal by standard reading, so it picks up signs the human eye misses. Second, it did even better in younger patients, with 95% sensitivity and 92% specificity. Results were similar in patients with and without a known gene mutation.
Key figures
- 0.96area under the curve in the test group (612 patients with HCM, 12,788 controls)
- 87% / 90%sensitivity and specificity in the test group
- 95% / 92%sensitivity and specificity in younger patients
- 0.95area under the curve in patients whose ECG looked normal
A thickened heart: getting ready for your cardiology visit with Mima
This AI is not in routine use yet: diagnosis still relies on an echocardiogram. But if a relative has hypertrophic cardiomyopathy, or your ECG raised a question, the study shows why it is worth raising with a cardiologist.
Mima helps you note shortness of breath, palpitations, or dizzy spells over time, and keep your ECG reports in one place. She also gets your question ready: "Given my family history, should I have an echocardiogram?"
What could this mean for patients?
An ECG is quick, painless, and available almost everywhere, from a primary care office to the ER. If an AI could flag possible HCM on it, people who don't know they have the disease could be sent sooner for an echocardiogram and follow-up care.
The authors highlight the strong results in younger patients. They describe a promising path toward HCM screening, not a tool ready for routine use.
What are the limits of this study?
The authors themselves say the model needs further refinement and external validation, meaning testing in other hospitals and populations.
In the study, people with HCM made up a much larger share of the test group than in real life, where the disease is uncommon. In real-world screening, the share of false alarms would therefore be higher. Finally, the study does not show that the tool improves patients' health.
Frequently asked questions
What is hypertrophic cardiomyopathy?
It is a disease in which the heart muscle becomes abnormally thick. It is often inherited and can go unnoticed for years.
Can an ECG detect hypertrophic cardiomyopathy?
Sometimes, but the ECG can also look normal. In this study, the AI flagged HCM even on ECGs read as normal, with an area under the curve of 0.95.
How is HCM diagnosed?
Doctors usually confirm HCM with an echocardiogram, sometimes followed by a cardiac MRI. The AI in this study only flags a suspicion.
Should my family be screened for HCM?
Because HCM often runs in families, the question comes up when a relative is diagnosed. A cardiologist can tell you who should be checked and how.
The original study
Ko WY, Siontis KC, Attia ZI, et al. “Detection of Hypertrophic Cardiomyopathy Using a Convolutional Neural Network-Enabled Electrocardiogram” Journal of the American College of Cardiology, February 1, 2020. DOI : 10.1016/j.jacc.2019.12.030
Read the full study in the Journal of the American College of CardiologyThis article summarizes a scientific study and is not medical advice.