AI could help sort heart ultrasounds faster by reading routine ECGs
PRESENT-SHD
The name of the AI model tested in the study.
triage
Choosing who should receive a test first.
area under the curve
A score for how well a model sorts people.
What happened
Researchers tested PRESENT-SHD, an AI model that estimates structural heart disease likelihood from 12-lead ECG images. Structural heart disease means problems in the heart’s muscle, valves, chambers, or function. The data came from a community cardiovascular screening program in Brazil.
Why the test matters
An electrocardiogram, or ECG, records the heart’s electrical activity. It is widely available and quick to perform. An echocardiogram uses ultrasound to examine heart structure and movement. It needs equipment, trained staff, and a scheduled slot. In places with limited imaging, people may wait while the queue grows.
The study asked whether an ECG could help decide who receives an echocardiogram first. The AI was used as a triage tool. It was not used as a final diagnosis.
What the study found
In one screening group, 3,282 adults had ECGs and point-of-care ultrasound. The AI separated people with probable structural heart disease better than traditional major-ECG-abnormality rules. Its area under the curve was 0.73, compared with 0.52 for the traditional method. This score measures ranking ability. It does not mean that 73 percent of people were diagnosed correctly.
The study also reported a 35 percent net reclassification improvement. Most of that gain came from identifying people unlikely to have disease. That matters because unnecessary referrals can consume scarce ultrasound capacity.
A second group included 1,475 adults with ECGs and full echocardiograms. Researchers simulated a service that could perform 200 echocardiograms monthly. To find 25 percent of structural heart disease cases, standard referral took 46 days. AI prioritization took six days. To find half the cases, it took 81 days versus 19 days. To find 90 percent, it took 164 days versus 96 days.
Those numbers came from a workflow simulation. They were not measured after changing a real hospital service. The original study Full text
Why it could matter
The proposed benefit is better use of a limited test. ECGs can be collected in more places than echocardiograms. An AI score could move higher-risk people forward. This could make diagnosis more timely in under-resourced settings. It could also help staff manage a long queue.
The model does not replace the echocardiogram. A high score still needs confirmation. A low score may also miss disease. The best role is a first sorting step, alongside clinical judgment.
What remains uncertain
The data came from one Brazilian screening program. Other countries may have different disease rates, equipment, and referral systems. Performance can change when a model meets a new population. The study used 12-lead ECG images. It did not show that consumer wearables would work the same way.
Most importantly, the simulation did not test whether AI-led triage reduces deaths, improves treatment, or prevents hospitalizations. A real-world study must measure missed cases, false alarms, waiting times, staff workload, and patient outcomes.
What to watch next
The next step is prospective testing in real clinics. Researchers will need local calibration and clear rules for referral. They should also compare results across age, sex, and other patient groups. The central question is practical: can AI help scarce heart imaging reach the right patients sooner, without creating a new queue of unnecessary tests?
AI reads ECGs to help choose who gets a heart ultrasound first
📰 Full story: AI could help sort heart ultrasounds faster by reading routine ECGs
A Brazilian study tested AI as a sorting helper. It did not use AI as a final diagnosis.
PRESENT-SHD
The name of the AI model used in the study.
echocardiogram
A heart test that uses sound waves.
simulation
A computer test of what might happen.
💡 The gist
- AI read ECGs and ranked people for heart ultrasounds.
- The test came from a Brazilian screening program.
- A simulation found cases sooner with AI sorting.
An electrocardiogram, or ECG, records the heart’s electrical activity. An echocardiogram uses sound waves to show the heart’s shape and movement. Doctors use the second test to confirm many structural heart problems.
The study tested PRESENT-SHD, an AI model that reads 12-lead ECG images. It looked for signs that someone might have structural heart disease. That means a problem with the heart’s muscle, valves, chambers, or pumping.
Why use an ECG first? ECGs are common and quick. Echocardiograms need special equipment and trained staff. A clinic may have fewer echo appointments than people waiting. An AI system could rank people who seem more likely to need one.
The researchers studied 3,282 adults with ECGs and a quick ultrasound. The AI’s sorting score was 0.73. Traditional ECG rules scored 0.52. A higher score means better separation. It does not mean 73 percent of patients had disease.
The researchers then used 1,475 people with ECGs and full echocardiograms. They imagined a service with 200 echo tests each month. Finding half the disease cases took 81 days with normal ordering. It took 19 days with AI prioritization. The study also found faster results for smaller and larger portions of cases.
The reason is simple. The AI does not create more appointments. It changes who gets an earlier appointment. That can matter when the queue is long.
The result is promising, but limited. The 19-day result came from a simulation. The researchers did not change a real clinic’s queue. The study also came from Brazil. Another country may have different patients and equipment.
A high AI score is not a diagnosis. Doctors still need an echocardiogram. A low score cannot guarantee that everything is normal. Researchers must test the system in real clinics. They must measure missed disease, extra scans, waiting times, and patient health.
This study supports AI as a sorting helper. It does not support replacing doctors or heart ultrasound. Read the original study
AI helped choose who gets a heart picture first
📰 Full story: AI could help sort heart ultrasounds faster by reading routine ECGs
A computer read heart lines. It helped plan the next test for some people.
PRESENT-SHD
The name of the computer system in the study.
ECG
A test that draws the heart’s electrical signals.
echocardiogram
A heart picture made with sound waves.
An ECG is a test that draws the heart’s tiny electrical signals. PRESENT-SHD is an AI system that reads those lines. AI means a computer that learns patterns from many examples. It looked for signs of heart problems.
An echocardiogram makes a heart picture with sound waves. Some doctors’ places have too few heart pictures. The study asked who should get one first. It used records from Brazil.
The computer put people in an order. It put people with more signs first. The study’s calculation found half the cases in 19 days. The usual order took 81 days.
These were computer-made results. A real hospital did not change its waiting order. A doctor still needs the heart picture. The computer cannot make the final decision. Other places may get different results. The study did not prove better health. It showed one way to choose the next test. Full study