Research Mentor Name

Dr. Domenico Gatti

Research Mentor Email Address

Institution / Department

Wayne State University School of Medicine

Document Type

Research Abstract

Research Type


Level of Research



Background: The 12-lead technique is the standard in ECG, however alternate cardiography modalities such as vectorcardiography (VCG) exist . While the VCG modality offers unique clinical metrics and certain advantages over ECG, it is hardly utilized due to it being more difficult to obtain than ECG. Here we introduce Cardio-Net, a MATLAB-based software that uses standard 12-lead ECG data to generate and visualize VCGs. Furthermore, we demonstrate the diagnostic potential of VCG by utilizing a recurrent neural network (RNN) to accurately classify vectorcardiograms.

Methods: MATLAB version 2019b and the following toolboxes were used for data processing: Deep learning, Wavelet, Signal processing, and Curve Fitting. ECGs were obtained from the Physionet ECG databases: CiPA ECG Validation Study. 128 patient ECGs (39 on placebo, 44 on chloroquine, 45 on verapamil) were used as training data for the classification of 32 more ECGs (8 on placebo, 16 on chloroquine, 8 on verapamil). A confusion matrix was then generated to validate the accuracy of the RNN in predicting the corresponding treatment group each VCG belonged to.

Results: Cardio-Net was successfully able to convert ECG into VCG that can be displayed in 3D space. It was also able to generate the QRS-T angle, a diagnostically-relevant parameter of VCG. The RNN-based classification of the VCG validation datasets operated at an overall accuracy of 96.88%.

Conclusion: We have demonstrated the ability of Cardio-Net to generate VCGs from ECG datasets and the potential of VCG for use in machine learning-based electrocardiogram classification.


Biomedical Engineering and Bioengineering | Cardiology | Computational Engineering | Medicine and Health Sciences


Thank you to Dr. Gatti for his mentorship and guidance throughout this project.