31Jul2025
09:00 Master's Defense Room 85 of IC2
Topic on
Filtering of ECG motion artifacts in free-standing conditions and detection of hyperglycemia
Student
Giorgio de Moraes Rossa
Advisor / Teacher
Marcelo da Silva Reis - Co-supervisor: Emely Pujólli da Silva
Brief summary
Studies have shown that the electrocardiogram (ECG) signal is capable of detecting important information from the heart, which can be used for various applications and analyses. One of these is the identification of changes in blood glucose. Currently, the methods of measuring blood glucose are invasive and can result in erythema and itching. The development of wearable sensors has led to the creation of different areas of research with data in free conditions, including ECG sensors, which can be used to improve the quality of life of patients who need to monitor their glucose continuously. However, one of the problems of the ECG signal is its vulnerability to noise and artifacts. Such noises are usually separated into baseline drift, power line interference, muscle artifacts and motion artifacts. Of these, electrode motion artifacts (EMs) are particularly difficult to remove, due to their similarity to normal ECG sinus rhythms. In uncontrolled situations, EM artifacts are more frequent and can disturb the signal. Despite the extensive research on ECG filtering, there is a lack of methodological standards, making direct comparisons between studies a challenging task. In this work, Extended Kalman Filter with weights and Denoising Autoencoders were used to remove motion artifacts from the ECG signal. This work proposes a methodology to validate the filtering techniques using different datasets, in a cross-dataset experiment, separating participants into training, validation, and testing, and mixing the input with clean segments and with real artifacts. The artifacts were also divided into training, validation, and testing, so as not to bias the model results. Finally, we apply the methods to filter the ECG from the D1NAMO dataset, which contains ECG and glucose captured in free conditions, to identify hyperglycemia events using HRV features of the ECG.
Examination Board
Headlines:
| Marcelo da Silva Reis | IC / UNICAMP |
| Fernando José Von Zuben | FEEC / UNICAMP |
| Marcos Medeiros Raimundo | IC / UNICAMP |
Substitutes:
| André Santanchè | IC / UNICAMP |
| Fabrício Martins Lopes | DACOM/UTFPR |