Researchers at Mayo Clinic have developed and externally validated an AI model that can identify potentially significant heart obstruction using standard ultrasound videos. The technology could help clinicians detect patients with hypertrophic cardiomyopathy (HCM) who require further examination, without initially relying on specialized Doppler imaging.
HCM is a genetic condition in which the heart muscle becomes abnormally thick. Around two thirds of patients develop left ventricular outflow tract (LVOT) obstruction, which restricts blood flow from the heart. This can cause symptoms including chest pain and shortness of breath during exercise or when lying flat. Identifying the obstruction is important because it can influence treatment decisions and long term management.
Currently, measuring LVOT obstruction generally requires Doppler echocardiography. According to senior author Imon Banerjee, an AI researcher at Mayo Clinic in Phoenix, this technique depends on precise alignment of the ultrasound beam and sufficient operator expertise. The researchers therefore investigated whether AI could identify subtle patterns in routinely acquired B mode ultrasound videos that clinicians cannot readily see.
AI analyzes three standard views
The study included 1,833 patients from Mayo Clinic. The model was tested in 275 patients and subsequently externally validated using data from 46 patients at a hospital in South Korea. Importantly, the AI system used only resting, non Doppler ultrasound videos. By analyzing these images, it predicted whether patients had a potentially significant obstruction affecting blood flow from the heart.
Combining information from three standard ultrasound views improved the model's ability to distinguish patients with elevated LVOT gradients. The technology also showed potential for identifying obstruction that may only become apparent when the heart is placed under stress. Despite substantial differences between the South Korean validation group and the population used to develop the algorithm, the model maintained strong performance. According to the researchers, this supports further investigation across different populations and clinical environments.
In a subset of cases, the AI model identified obstruction more accurately than two expert echocardiographers who evaluated the same non Doppler images. This finding also illustrates the difficulty of detecting LVOT obstruction from conventional two dimensional ultrasound images without Doppler measurements.
Supporting earlier diagnosis
The researchers stress that the AI technology is intended to complement rather than replace Doppler echocardiography. Its potential value lies in identifying patients who should receive additional diagnostic testing. An AI based assessment of routine ultrasound images could prompt clinicians to perform confirmatory Doppler measurements, stress testing or refer patients to a specialist HCM center. This could be particularly relevant in healthcare settings where comprehensive echocardiography expertise is not readily available.
The approach could also expand the possibilities of portable ultrasound. If potentially significant LVOT obstruction can be flagged from standard imaging, patients could potentially be screened in settings without immediate access to advanced Doppler examinations. Further research is needed before the technology can be implemented clinically. The Mayo Clinic team plans prospective validation across broader patient populations, clinical settings and different ultrasound platforms.
Portable AI-drive ultrasound
Last year, Philips introduced the Compact Ultrasound 5500CV, a portable ultrasound system designed to accelerate and improve access to cardiac imaging using AI. Suitable for bedside use in hospitals and other care settings, the system can reduce scan times by up to 50 percent through AI assisted measurements, according to Philips.
Its Auto Measure functionality automates routine 2D and Doppler cardiac measurements, supporting greater consistency and reproducibility. The system also supports advanced transducers for transesophageal echocardiography and uses xPlane Imaging to capture two image planes simultaneously. Compatibility with the S12 4 transducer extends its use to neonatal and pediatric care.
The Compact Ultrasound 5500CV integrates with Philips’ existing ultrasound platforms and Ultrasound Workspace. Through Collaboration Live, healthcare professionals can also receive real time remote diagnostic support, potentially supporting hybrid care and access to specialist expertise.
References
Circulation: Cardiovascular Imaging (research)
