MAIOR NETO, L. A. S.; MAIOR NETO, Luis Alberto Souto.
Résumé:
Electrocardiography (ECG) is a kind of medical exam that makes use of electric signals
captures from a patient’s skin. Physicians diagnose cardiac diseases most often using ECG,
although ECG diagnosis is known to be time consuming and to require strong medical
expertise. At northeastern Brazil, in inland cities, there is a lack of medical professionals
to supply the healthcare demand on ECG examinations. In emergency situations, the lack
of medical specialists can lead to delay in proper clinical examination and, thus, proper
diagnosis and treatment, putting patient lifes in risk. In order to solve this issue, in this
work accomplished at NUTES - UEPB as part of a mandatory undergraduate internship,
we propose a tool named Intelligent Detection of Arrhythmic Heartbeats on ECG, or
IDAH-ECG, which consists of a system comprised of modules for portable ECG exam
acquisition, server monitoring, cloud-based signal processing, and short message service
(SMS) notification. The system makes use of a portable ECG device to capture the ECG
signal during an examination, uploads the exams to a server, which proceeds to process
the signals and to distinguish arrhythmic heartbeats from healthy, or normal ones. Upon
detection of any abnormalities, a SMS is sent to available medical professionals in the
area, which then can use the message to access the cloud-based database of examinations,
evaluate the acquired examination and then promptly decide how to proceed with the
patient. This document focuses on the signal processing aspect of the system which makes
use of Finite Impulse Response and Discrete Wavelet Filtering, Principal Component
Analysis and Artificial Neural Networks to accomplish arrhythmia detection with accuracy
of 96.48%, sensitivity of 98.7%, using ten-fold cross-validation and Monte Carlo testing
over the MIT-BIH Arrhythmia Database.