Scholarlink

<p>The fetal electrocardiogram is crucial for fetal arrhythmia detection. There are currently very few guidelines for interpreting fetal ECG signals(fECG), and well it contains noise such as mother ECG (MECG), uterine contraction, electromyogram, and various fetal orientations inside. This study proposes a deep learning structure, ‘fECG-net’ 2-dimensional CNN, and an efficient preprocessing technique for fetal arrhythmia detection. Method: Abdominal ECG(AECG) is transformed into time-scale images using the continuous wavelet transform. These images are then fed as input data to the fECG-net CNN classifier. fECG-net 2D CNN detected fetal arrhythmia using the Physionet Noninvasive Fetal Arrhythmia database(NIFEA DB). Experiments are conducted on single and multiple-channel abdominal signal recordings. The performance of the proposed fECG-net CNN model is compared with three pre-trained models, Alexnet, Resnet50, and Vgg16, through a transfer learning approach. As a result, we achieved 97.5% accuracy for the proposed CNN model and accuracy of 92.98%, 87.5%, and 90% for the Alexnet, Resnet50, and Vgg16 CNN models, respectively. The proposed preprocessing and fECG-net CNN achieve excellent classification accuracy in their state-of-the-art, computationally efficient way. Therefore, the fECG-net CNN best suits for DL-based noninvasive fetal monitor devices for fetal cardiac disease detection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

fECG-net: Fetal ECG Arrhythmia Detection CNN from Abdominal ECG Recordings

  • 期刊论文

  • 发表日期：

    2025年8月4日

查看全文
  • Vinita Yerande,
  • Kalyani Bhole,
  • D. N. Sonawane,
  • C. Y. Patil

摘要

The fetal electrocardiogram is crucial for fetal arrhythmia detection. There are currently very few guidelines for interpreting fetal ECG signals(fECG), and well it contains noise such as mother ECG (MECG), uterine contraction, electromyogram, and various fetal orientations inside. This study proposes a deep learning structure, ‘fECG-net’ 2-dimensional CNN, and an efficient preprocessing technique for fetal arrhythmia detection. Method: Abdominal ECG(AECG) is transformed into time-scale images using the continuous wavelet transform. These images are then fed as input data to the fECG-net CNN classifier. fECG-net 2D CNN detected fetal arrhythmia using the Physionet Noninvasive Fetal Arrhythmia database(NIFEA DB). Experiments are conducted on single and multiple-channel abdominal signal recordings. The performance of the proposed fECG-net CNN model is compared with three pre-trained models, Alexnet, Resnet50, and Vgg16, through a transfer learning approach. As a result, we achieved 97.5% accuracy for the proposed CNN model and accuracy of 92.98%, 87.5%, and 90% for the Alexnet, Resnet50, and Vgg16 CNN models, respectively. The proposed preprocessing and fECG-net CNN achieve excellent classification accuracy in their state-of-the-art, computationally efficient way. Therefore, the fECG-net CNN best suits for DL-based noninvasive fetal monitor devices for fetal cardiac disease detection.