Publicação
Classificação de sinais de ECG com técnicas explicáveis de inteligência artificial
| Resumo: | Cardiovascular diseases are among the main causes of premature mortality, with atrial flutter representing a clinically relevant arrhythmia due to its association with stroke and heart failure. The eletrocardiogram (ECG) is the most suitable diagnostic method for evaluating chardiac rhytms. Automatic ECG interpretations have attempted to improve clinical practice. However, the lack of interpretability of existing models has limited their acceptance. This dissertation presents a framework for atrial flutter classification using raw 12- lead ECG signals from PTB-XL Database, Georgia 12-Lead ECG Challenge Database and Large 12-Lead ECG Database for Arrhythmia Study. A hybrid deep learning model combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks was developed and trained under a 10-fold cross-validation scheme. To ensure model transparency, explainable artificial intelligence (XAI) mehods were applied: Shapley Additive Explanations (SHAP) was used to quantify the contribution of each lead, and Local Interpretable Model-Agnostic Explanations (LIME) was employed to highlight the most informative temporal segments at the patient level. The results demonstrate that the proposed approach achieves competitive performance while improving interpretability, thus contributing to more reliable and clinically meaningful applications of artificial intelligence in cardiology. |
|---|---|
| Autores principais: | Martins, Hugo Fidalgo Oliveira |
| Assunto: | Eletrocardiogram (ECG) Atrial flutter (AFL) Deep learning (DL) Ex- plainable AI (XAI) |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Instituto Politécnico de Bragança |
| Idioma: | português |
| Origem: | Biblioteca Digital do IPB |
| Resumo: | Cardiovascular diseases are among the main causes of premature mortality, with atrial flutter representing a clinically relevant arrhythmia due to its association with stroke and heart failure. The eletrocardiogram (ECG) is the most suitable diagnostic method for evaluating chardiac rhytms. Automatic ECG interpretations have attempted to improve clinical practice. However, the lack of interpretability of existing models has limited their acceptance. This dissertation presents a framework for atrial flutter classification using raw 12- lead ECG signals from PTB-XL Database, Georgia 12-Lead ECG Challenge Database and Large 12-Lead ECG Database for Arrhythmia Study. A hybrid deep learning model combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks was developed and trained under a 10-fold cross-validation scheme. To ensure model transparency, explainable artificial intelligence (XAI) mehods were applied: Shapley Additive Explanations (SHAP) was used to quantify the contribution of each lead, and Local Interpretable Model-Agnostic Explanations (LIME) was employed to highlight the most informative temporal segments at the patient level. The results demonstrate that the proposed approach achieves competitive performance while improving interpretability, thus contributing to more reliable and clinically meaningful applications of artificial intelligence in cardiology. |
|---|