Electrocardiogram-compatible deep learning algorithm for aortic stenosis detection

This technology is a diagnostic tool combining electrocardiography with the power of deep learning algorithms to detect moderate to severe aortic stenosis, aortic regurgitation, and mitral regurgitation..

Unmet Need: Early detection of valvular heart disease and interventive treatment

If left untreated, valvular heart disease can lead to serious complications including heart failure and death. Yet, most patients do not seek treatment until symptoms appear, which often occurs late in disease progression when interventions are less effective. As a result, many patients remain undiagnosed and fail to receive early interventional valve replacements. Currently, there are no cost-appropriate, population-wide screening practices to detect early signs of valvular heart disease, which may ultimately lead to better patient management and improved outcomes.

The Technology: Automated, universal screening tool leveraging lost-cost electrocardiogram

This technology uses deep learning analysis of the 12 lead electrocardiogram to accurately diagnoses aortic stenosis, aortic regurgitation, and mitral regurgitation. Rather than relying on patient presentation of symptoms and subsequent echocardiograms, this platform feeds low-cost electrocardiogram measurements to a trained deep learning algorithm which outputs patient diagnosis. Therefore, this technology offers an accessible solution to patient-wide screening, enabling earlier intervention and improved patient outcomes.

This technology was validated against echocardiography diagnosis in an independent, prospective multicenter study.

Applications:

• Universal screening tool for the detection of aortic stenosis, aortic regurgitation, and mitral valve regurgitation
• Research tool for understanding electrocardiogram characteristics of valvular heart disease
• Indication for early intervention, and increased number of effective valve replacements

Advantages:

• Irrespective of patient symptom presentation
• Early identification of disease
• Cost-effective
• Increases hospital revenue by increasing number of electrocardiograms
• Reduces mortality rate and improves hospital outcomes with early detection
• Uses proprietary model weights

Lead Inventor:

Timothy Poterucha, M.D.

Patent Information:

Patent Pending

Related Publications:

• Poterucha TJ, Elias P, Hahn RT, Nazif T, Maurer MS, Einstein AJ, Leon MB, Perotte A, Cohen AJ “Cost effectiveness of screening for valvular heart disease using valvenet deep learning electrocardiogram model” Circulation. 2021 Nov 8; 144(A9735).

• Poterucha TJ, Elias P, Bhave S, Rodriguez V, Einstein A, Perotte A, Maurer M. “Deep learning analysis of cardiac testing for the detection of cardiac amyloidosis” J Am Coll Cardiol. 2021 May 15; 77(18Supplement1): 529.

Tech Ventures Reference:

• IR CU21335

• Licensing Contact: Sara Gusik