Detection and localization of myocardial infarction using vectorcardiography
A method includes detecting whether one or more myocardial infarctions (MI) has occurred using vectorcardiographic (VCG) signals with gradient boosting, the VCG signals including VCG loops, and determining an MI location using the VCG signals and gradient boosting.
1. A method comprising:
detecting electrocardiogram (ECG) signals from a patient;
determining vectorcardiographic (VCG) signals as a function of the ECG signals, wherein the VCG signals comprise VCG loops;
extracting features from the VCG signals, wherein the extracted features comprise at least one of geometrical features of the VCG loops and/or spatio-temporal information of the VCG loops;
detecting, using one or more trained machine learning models, one or more myocardial infarctions (MI) based on the extracted features; and
determining, using the one or more trained machine learning models, an MI location of the detected one or more myocardial infarctions based on the extracted features.
2. The method of claim 1 , wherein determining the MI location includes identifying a region of a heart of where the one or more MI has occurred.
3. The method of claim 1 , wherein the geometrical features include one or more of loop perimeter, loop centroid, maximum vector length (MVL), loop area, Ratio Perimeter/Area, Maximum distance between centroid and VCG loop, Angle between MVL and the different planes, Angle between QRS and T wave maximum vectors, or Angle between QRS and T wave optimal planes.
4. The method of claim 1 , wherein the geometrical features of the VCG loops includes geometrical features of QRS VCG loops and T-wave VCG loops.
5. The method of claim 1 , wherein the spatio-temporal information includes one or more of octant average vector length, maximum vector length (MVL) per octant, percentage of time per octant, or variance of vector magnitude per octant.
6. The method of claim 1 , wherein the spatio-temporal information of the VCG loops includes spatio-temporal distribution information of the VCG loops for both QRS VCG loops and T-wave VCG loops.
7. The method of claim 1 , wherein the one or more trained machine learning models is trained using a gradient boosting method.
8. A system comprising:
a set of electrodes configured to sense electrocardiogram (ECG) signals; and
processing circuitry configured to:
determine vectorcardiographic (VCG) signals as a function of the ECG signals, wherein the VCG signal comprise CCG loops;
extract features from the VCG signals, wherein the extracted features comprise at least one of geometrical features of the VCG loops and/or spatio-temporal information of the VCG loops;
detect, using one or more trained machine learning models, one or more myocardial infarctions (MI) based on the extracted features; and
determine, using the one or more trained machine learning models, a MI location of the detected one or more myocardial infarctions based on the extracted features using the VCG signals and gradient boosting.
9. The system of claim 8 , wherein the determination of the MI location includes identification of a region of a heart of where the one or more MI has occurred.
10. The system of claim 8 , wherein the geometrical features include geometrical features of QRS VCG loops and T-wave VCG loops.
11. The system of claim 8 , wherein the geometrical features include one or more of loop perimeter, loop centroid, maximum vector length, loop area, Ratio Perimeter/Area, Maximum distance between centroid and VCG loop, Angle between MVL and the different planes, Angle between QRS and T wave maximum vectors, or Angle between QRS and T wave optimal planes.
12. The system of claim 8 , wherein the spatio-temporal information includes one or more of octant average vector length, maximum vector length (MVL) per octant, percentage of time per octant, or variance of vector magnitude per octant.
13. The system of claim 8 , wherein the spatio-temporal information of the VCG loops includes spatio-temporal distribution information of the VCG loops for both QRS VCG loops and T-wave VCG loops.
14. The system of claim 8 , wherein the one or more trained machine learning models is trained using a gradient boosting method.
15. A non-transitory computer-readable storage medium comprising instructions, that when executed, cause processing circuitry to:
receive electrocardiogram (ECG) signals from a patient;
determine vectorcardiographic (VCG) signals as a function of the ECG signals wherein the VCG signals comprise VCG loops;
extract features from the VCG signals wherein the extracted features comprise at least one of geometrical features of the VCG loops and/or spatio-temporal information of the VCG loops;
detect, using one or more trained machine learning models, one or more myocardial infarctions (MI) based on the extracted features; and
determine, using the one or more trained machine learning models, an MI location of the detected one or more myocardial infarctions based on the extracted features.
16. The non-transitory computer-readable storage medium of claim 15 , wherein the one or more trained machine learning models is trained using a gradient boosting method.