Non-invasive fetal electrocardiogram (FECG) is used to monitor the electrical pulse of the fetal heart. Decomposing the FECG signal from maternal ECG (MECG) is a blind source separation problem, which is hard due to the low amplitude of FECG, the overlap of R waves, and the potential exposure to noise from different sources. Traditional decomposition techniques, such as adaptive filters, require tuning, alignment, or preconfiguration, such as modeling the noise or desired signal to map MECG to FECG efficiently. This novel architecture based on the attention layer, sine activation function and cycle generative adversarial neural network is utilized to map maternal and fetal ECG. The proposed method is evaluated in two forms. First, the quality of FECG extracted from MECG is evaluated. Second, the fetal QRS detection from MECG is assessed. On the abdominal and direct FECG (A&D FECG) dataset, an average 98% R-Square [CI 95%: 97%, 99%] as the goodness of fit and 99.7 % F1-score [CI 95%: 97.8%, 99.9%] for QRS estimation achieved based on subject-leave-out validation. Besides A&D FECG dataset, the non-invasive FECG (NI-FECG) and NI-FECG challenge datasets are also used for fetal QRS estimation and achieved 99.7 % F1-score [CI 95%: 97.8-99.9], 99.6% F1-score [CI 95%: 98.2%, 99.9%] and 99.3% F1-score [CI 95%: 95.3%, 99.9%], respectively.
Input variables : Mother ECG
Output Variables : Fetal Ecg
Visit Model : github.com
Additional links : arxiv.org
| Model Category | : | Public |
| Date Published | : | January, 2021 |
| Healthcare Domain | : | Medical Technology |
| Code | : | github.com |
| Data Privacy |
| Synthetic Data Generation |