AI model detects heart obstruction from routine ultrasound, outperforms expert echocardiographers

AI model identifies heart obstruction without specialized imaging

Mayo Clinic researchers developed an AI model that detects left ventricular outflow tract obstruction in hypertrophic cardiomyopathy patients using only routine ultrasound videos, eliminating the need for specialized Doppler echocardiography1

1

. Published in Circulation: Cardiovascular Imaging, this technology addresses a critical gap in cardiac diagnostics by identifying subtle patterns imperceptible to the human eye in B-mode ultrasound images2

2

. The model trained on 1,833 patients in the Mayo Clinic cohort, tested on 275 patients, and externally validated in 46 patients from a South Korean hospital, demonstrating robust performance across different populations1

1

.

Addressing challenges in hypertrophic cardiomyopathy diagnosis

Hypertrophic cardiomyopathy affects approximately two-thirds of patients who develop left ventricular outflow tract (LVOT) obstruction, restricting blood flow from the heart and causing chest pain and shortness of breath2

2

. Traditional diagnosis requires Doppler echocardiography, which depends on precise ultrasound-beam alignment and operator expertise, creating barriers in resource-limited settings where specialized echocardiography expertise may not be readily available1

1

. Senior author Imon Banerjee, Ph.D., an AI researcher at Mayo Clinic in Phoenix, explained the team wanted to determine whether AI-assisted cardiac diagnostics could recognize patterns in routinely acquired ultrasound videos to enable earlier diagnosis and timely confirmatory Doppler evaluation2

2

.

Superior performance in detecting obstructions

The AI model achieved notable results by combining information from three standard ultrasound views, improving its ability to distinguish patients with elevated LVOT gradients1

1

. In a subset of cases, the model identified obstruction more accurately than two expert echocardiographers who reviewed the same non-Doppler images, highlighting how difficult recognizing LVOT obstruction can be from routine two-dimensional images without Doppler measurements2

2

. The technology also helped identify obstruction that may only appear when the heart is under stress, expanding diagnostic capabilities beyond resting conditions1

1

.

Implications for patient management and access

Dr. Banerjee emphasized the technology is intended to complement, not replace, Doppler echocardiography, enabling earlier identification of patients with potential LVOT obstruction and prompting confirmatory testing, stress testing, or referral to an HCM specialty center2

2

. The model maintained strong performance in the external validation despite substantial differences between the South Korean population and the development cohort, supporting broader clinical implementation across different patient populations and settings1

1

. This capability could expand access to earlier screening and risk assessment using portable ultrasound in environments where comprehensive Doppler assessment is not readily available, potentially transforming how clinicians approach hypertrophic cardiomyopathy detection2

2

. Next steps include additional prospective validation across broader clinical settings, ultrasound platforms, and patient populations to establish the model's utility in real-world practice1

1

.