This technology is an attention-based deep learning for cardiac right ventricular volume quantification using 2D echocardiography.
Unmet Need: Low-cost, widely-available tools for measurements of ventricle volume
To diagnose cardiac diseases and follow up response to treatment, doctors often need accurate measurements of the right ventricle, a core structure of the heart. However, the gold standard for right ventricular quantification, cardiovascular magnetic resonance imaging (CMR), is not widely available, whereas widely used imaging techniques such as two-dimensional transthoracic echocardiography (2DE) are typically inaccurate due to the complex geometry of the right ventricle.
The Technology: Deep learning model for accurate right ventricular volume quantification
This technology describes a deep learning method, namely an attention-based deep learning network, capable of using 2DE data for quantification of the right ventricle. Specifically, the method uses measurements performed by physicians on 2DE to calculate the volume and the function of the right ventricle with an accuracy that approximates that of CMR. This technology has been developed and tested with a retrospective study of 50 patients.
Applications:
- Clinical tool to diagnose cardiac abnormalities, specifically of the right ventricle
- Clinical tool to follow up response to treatment, specifically for abnormalities of the right ventricle
- Research tool for studying cardiac abnormalities, specifically of the right ventricle
- Training tool for clinicians learning echocardiography
Advantages:
- Based on widely-used technology of two-dimensional transthoracic echocardiography (2DE)
- Improved accuracy that approximates cardiovascular magnetic resonance imaging (CMR with less cost
- Easy workflow integration
Lead Inventor:
Polydoros Kampaktsis, M.D., Ph.D.
Patent Information:
Patent Pending
Related Publications:
Kampaktsis PN, Bohoran TA, Lebehn M, McLaughlin L, Leb J, Liu Z, Moustakidis S, Siouras A, Singh A, Hahn RT, McCann GP, Giannakidis A. An attention-based deep learning method for right ventricular quantification using 2D echocardiography: Feasibility and accuracy. Echocardiography. 2023.
Bohoran TA, Kampaktsis PN, McLaughlin L, Leb J, Moustakidis, S, McCann G, Giannakidis A. Right Ventricular Volume Prediction by Feature Tokenizer Transformer-based Regression of 2D Echocardiography Small-Scale Tabular Data. Functional Imaging and Modeling of the Heart 2023 Conference; 2023.
Bohoran TA, Kampaktsis PN, Leb J, Moustakidis S, McCann G, Giannakidis A. Embracing Uncertainty Flexibility: Harnessing a Supervised Tree Kernel to Empower Ensemble Modelling for 2D Echocardiography-Based Prediction of Right Ventricular Volume. The 16th International Conference on Machine Vision (ICMV 2023); 2023.
Tech Ventures Reference:
