Artificial intelligence could enable people without previous ultrasound experience to capture usable heart images and help identify patients who may have aortic stenosis. In a Mayo Clinic study, novice users performed focused cardiac ultrasound examinations after only four hours of training, supported by AI for both image acquisition and analysis.
The prospective study, presented at the 2026 ESC Congress, suggests that the combination could potentially expand access to screening in settings where comprehensive echocardiography and trained imaging specialists are less readily available.
Lowering the threshold for ultrasound
Aortic stenosis occurs when the aortic valve narrows, making it more difficult for the heart to pump blood through the body. The condition affects approximately 7 percent of people aged 75 and older and is the most common reason for heart valve intervention worldwide. Symptoms may not develop until the disease has progressed, making earlier detection potentially valuable for monitoring and treatment.
Comprehensive echocardiography is the standard diagnostic test, but requires specialised equipment and professionals trained in acquiring and interpreting ultrasound images. According to senior author Gal Tsaban, a cardiologist at Mayo Clinic, these requirements can limit availability, particularly in resource-constrained settings. They also make comprehensive echocardiography impractical as a broad screening method.
The researchers therefore investigated whether focused cardiac ultrasound combined with AI could lower the expertise required for initial screening. They first developed and validated a deep learning algorithm using echocardiograms from patients treated across several Mayo Clinic locations. The model was subsequently evaluated using handheld ultrasound images acquired by experienced sonographers.
Novices achieve high screening accuracy
For the prospective part of the study, nine research staff members without previous clinical or ultrasound experience received four hours of training. They then used AI guidance while performing focused cardiac ultrasound examinations.
The AI system could analyse nearly 97 percent of the examinations. It correctly identified 93 percent of patients who had moderate or more severe aortic stenosis and correctly ruled out the condition in 96 percent of patients who did not have it. Around 10 percent of examinations were flagged for assessment by a cardiac imaging specialist. Adding this human review reduced the number of false-positive results. However, the researchers found that the combined approach also resulted in some patients with aortic stenosis not being identified. The results demonstrate both the potential and limitations of AI-assisted screening. The technology can support inexperienced users in acquiring and interpreting cardiac images, but specialist involvement and further diagnostic testing remain necessary.
Screening rather than diagnosis
The researchers emphasise that the approach is not designed to replace comprehensive echocardiography or physician assessment. Its intended role is to identify people who may have clinically relevant aortic stenosis and could benefit from further evaluation. Patients flagged as potentially having moderate or severe disease would still require a comprehensive echocardiogram to confirm the diagnosis and establish its severity.
According to the researchers, this distinction is particularly important if AI-assisted handheld ultrasound is introduced in communities with limited access to specialised cardiovascular imaging. The technology could provide an additional first step in the diagnostic pathway rather than an alternative to specialist care.
The study suggests that combining relatively short training with AI-guided image acquisition and automated analysis could make cardiac screening accessible to a broader group of healthcare workers. Further research will be needed to establish how the approach performs when implemented across different clinical settings and patient populations.
Add ICT&health on Google
Show more content from ICT&health in Google Search.
