Optimizing Deep Learning Architectures with Convolutional Networks for Enhanced ECG Signal Classification: Addressing Class Imbalance and Hyperparameter Tuning Challenges

Artificial intelligence, through machine learning and deep learning (DL), is transforming fields like medical diagnosis, including ECG classification. Traditional methods, relying on manually crafted features, often fall short in complex tasks. Our study introduces a DL architecture using ۱-D convolution and Fully Convolutional Network (FCN) layers, inspired by VGGNet and ResNet, to improve ECG classification. We tackle issues such as imbalanced datasets and multiclass classification by weighting classes and employing confusion matrices and F۱-scores. Our approach demonstrates superior accuracy and efficiency, highlighting the transformative potential of DL in healthcare.

کلیدواژه ها:

Deep learning ، VGG۱۶ ، ResNet ، Electrocardiogram Signal ، multiclass-metrics