Desarrollo de un sistema de monitoreo cardíaco inteligente de bajo costo utilizando TinyML.

Abstract

The project develops a portable cardiac monitoring system based on TinyML to detect arrhythmias in an accessible and affordable way. The main objective is to improve diagnostic capacity in areas with limited access to specialized medical equipment. The hypothesis is that a low-cost device can perform continuous monitoring in real time, facilitating the early detection of cardiac conditions and reducing disparities in health care. The justification lies in the high incidence of cardiovascular diseases in Ecuador and the need for inclusive solutions. The system was built using an ESP32 microcontroller, an AD8232 sensor, and an ADS1115 converter, integrated with the Edge Impulse platform to train a machine learning model. ECG signals from the public database PTB-XL were used, complemented with additional data collected from patients in different age ranges. These signals were processed using digital filters to improve their quality and reduce noise. Subsequently, classification models were retrained to differentiate between normal heart rhythms and episodes of atrial fibrillation. The results obtained from the tests carried out in Edge Impulse used inference with a segment of the dataset used in which the model was retrained and thus showed a 95% efficiency in detecting arrhythmias, with an 88% effectiveness in minimizing false positives. The prototype proved to be functional and accessible, with a total cost of $56.50. It is concluded that the proposed system is a viable tool for real-time cardiac monitoring, contributing to the early detection of cardiovascular diseases and improving equity in access to medical services. Keywords: Cardiac monitoring, TinyML, arrhythmias, accessibility