Identification of Atypical Cardiac Patterns Before and After Exercising Using Artificial Intelligence and Eulerian Video Magnification

This project is part of the research line of emerging technologies in rehabilitation of the BASPI-FootLab research group of the Javeriana University. The project is focused on the design, implementation and verification of an intelligent system that is able to identify and report to the user about atypical values in the cardiac frequency before and after doing physical activity. To achieve this objective, an algorithm was implemented to detect the cardiac frequency from very small changes of colours on the skin face and training a program to detect atypical cardiac patterns. This project was developed in three phases. The first one consisted in the implementation of the Eulerian video magnification, this algorithm made it possible to detect the cardiac frequency of a person with an acceptable accuracy for our task. The second phase was the validation of the chosen method against a commercial pulse oximeter. Finally, the final phase consisted of the selection and coding of an artificial intelligence algorithm that was trained with cardiac patterns obtained from two different published databases and validated with our proprietary database. This algorithm seeks to predict whether the heart of a person shows any atypical cardiac pattern before or after a brisk walk. When the logistic regression algorithm was tested in a proprietary database, a precision of 75% was obtained. It seems that this innovative system may contribute to generating early warnings for cardiac problems. This development is still under study and will continue to be validated with a bigger group of volunteers.