Systems and methods for electromechanical wave imaging with machine learning for automated activation map generation for display on an image processing unit
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12642504-B2
Publication Date
2026-06-02
Expiration Date
Abstract
The present subject matter relates to techniques for electromechanical wave imaging. The disclosed system can include a processor that can be configured to perform an automated selection of at least one zero-crossing location using a heuristic-based baseline and/or a machine learning classifier and generate an electromechanical wave imaging isochrone based on the automated selection.
Core Innovation
The invention relates to electromechanical wave imaging of a tissue of a subject using ultrasound image data acquired by a transducer and received by a processor as a dataset. The processor performs an automated selection of at least one zero-crossing location within the dataset using a heuristic-based baseline and/or a machine learning classifier. Based on the automated selection, the processor generates an electromechanical wave imaging isochrone for display on an imaging processing unit.
A key aspect is that the zero-crossing location selection is automated, rather than selected manually, using either a heuristic-based baseline and/or a machine learning classifier. The selection and isochrone generation are performed by the processor from the ultrasound-derived dataset. The automated selection is used to produce electromechanical wave imaging isochrones for display on an imaging processing unit.
In implementations described in the provided content, the automated selection can use a Random Forest, logistic regression, and/or a support vector machine classifier, and may include a heuristic-based baseline. The selection may be driven by probabilistic outputs for zero-crossing locations, including selecting the highest-probability zero-crossing per pixel or applying a probability cutoff threshold tied to dataset sensitivity and/or true positive value. The selected zero-crossing locations are then used to generate activation isochrones and activation maps.
Claims Coverage
The document provides two independent claims (a system claim and a method claim) that share the core inventive concept of automated zero-crossing selection and electromechanical wave imaging isochrone generation. Across these independent claims, the main inventive features are automation of zero-crossing selection using a heuristic-based baseline and/or a machine learning classifier, followed by generation of an electromechanical wave imaging isochrone for display.
Automated selection of zero-crossing locations for electromechanical wave imaging isochrone generation
The processor performs an automated selection of at least one zero-crossing location within the dataset using a heuristic-based baseline and/or a machine learning classifier.
Isochrone generation for display based on automated selection
The processor generates an electromechanical wave imaging isochrone for display on an imaging processing unit based on the automated selection.
Receiving ultrasound image data dataset for automated processing
The transducer is configured to acquire ultrasound image data of the tissue constituting a dataset, and the processor receives the dataset.
Machine-implemented automated zero-crossing selection and isochrone generation in an electromechanical imaging method
The method includes performing, by the processor, an automated selection of at least one zero-crossing location using a heuristic-based baseline and/or a machine learning classifier, and generating, by the processor, an electromechanical wave imaging isochrone for display based on the automated selection.
Both independent claims cover an electromechanical wave imaging workflow in which ultrasound image data are acquired and provided as a dataset to a processor, the processor automatically selects at least one zero-crossing location using a heuristic-based baseline and/or a machine learning classifier, and the processor generates an electromechanical wave imaging isochrone for display based on that automated selection.
Stated Advantages
Reduced processing time versus manual processing (e.g., <5 minutes vs ~90 minutes).
Maintained activation propagation patterns when applied to reduced ultrasound frame rates (500/250/125 Hz).
Robust performance of the Random Forest classifier across described evaluation settings (including canines and WPW/sinus rhythm patients).
Use of probability voting/probability cutoff to support semi-automated correction aligned with probability-based rules (including precision-recall evaluation).
Documented Applications
Generation of electromechanical activation isochrones/activation maps in described patient contexts, including canines and WPW/sinus rhythm patients.
Application to a Wolff-Parkinson-White patient, an Atrial Flutter patient, or a ventricular tachycardia patient.
Application to reduced ultrasound frame rates (500/250/125 Hz) while maintaining activation propagation patterns and similarity metrics.
Semi-automated manual correction using probability thresholding tied to selection rules.
Interested in licensing this patent?