Personalized chest acceleration derived prediction of cardiovascular abnormailities using deep learning

NU 2023-121

INVENTORS

Michael Markl *

Ethan M.I. Johnson

Mahmoud Ebrahimkhani

SHORT DESCRIPTION

A system and method for predicting cardiovascular function using machine learning from physiological patient data.

BACKGROUND

Cardiac and flow abnormalities can often go undetected until serious health problems arise. Diagnosing these abnormalities often requires expensive medical imaging techniques like MRI. Seismocardiography (SCG), which measures mechanical forces from the heart, offers a potential low-cost alternative. SCG can provide valuable insights into cardiac function, especially when its signal features are paired with powerful machine learning techniques to identify subtle changes. While MRI remains the gold standard for cardiac assessment, SCG could play a valuable role in initial screening and monitoring.

ABSTRACT

This technology introduces a sophisticated approach to predict cardiovascular function by leveraging machine learning techniques. Utilizing a combination of physiological measurement data and patient data, the system employs a neural network trained on extensive datasets to accurately quantify cardiovascular functions. This method can be used both to identify aortic valve status/dysfunction and to predict key clinical metrics, such as peak systolic velocity, enhancing patient care through precise cardiovascular assessments.

APPLICATIONS

• Non-invasive cardiovascular diagnostics.
• Healthcare data analysis platforms.
• Personalized patient care and monitoring.
• Development of advanced healthcare predictive models.
• Integration into clinical decision support systems.

ADVANTAGES

• Cost-effective and efficient compared to traditional imaging methods.
• Non-invasive with minimal requirement for skilled operators.
• Capable of providing accurate cardiovascular function metrics.
• Facilitates the diagnosis of cardiovascular abnormalities.
• Utilizes advanced machine learning models for precise predictions.

PUBLICATION

Ann Biomed Eng 51, 2802–2811 (2023)

IP STATUS

US Patent Pending

INVO CONTACT