Enhancing fetal electrocardiogram classification: A hybrid approach incorporating multimodal data fusion and advanced deep learning models

The accurate identification of combined fetal and maternal electrocardiogram (ECG) signals plays a crucial role in assessing the health and well-being of both the pregnant woman and the fetus. This paper presents a novel framework that addresses key challenges in fetal ECG identification, including reducing the number of inputs/observations and optimizing the number of iterations required for reliable signal detection. The proposed approach leverages a combination of principal component analysis (PCA) and independent component analysis (ICA) on cardiological data obtained from implanted electrodes on the central abdomen of pregnant women.To mitigate the potential adverse effects associated with a large number of electrodes, which can impact maternal and fetal health, this framework emphasizes the importance of controlling the number of inputs. Furthermore, excessive algorithmic complexity is managed by optimizing the number of iterations. Classical statistical methods often exhibit redundancy, thus posing a significant challenge. To overcome this, an intelligent machine is introduced that evaluates the quality of estimated sources or extracted signals through time-scale images classification, employing the specially designed AlexNet Deep Neural Network Architecture. By utilizing minimal inputs, even as low as two, this intelligent machine assesses the performance of the model on test data, employing metrics such as accuracy, precision, recall, and F1 score. These metrics provide valuable insights into the blind separation model’s capacity and enable the determination of the optimal number of iterations and inputs, or the simultaneous adjustment of both parameters. The proposed framework offers a comprehensive solution to enhance fetal ECG identification, promoting improved healthcare monitoring and diagnosis during pregnancy.