top of page

How Deep learning can be used to extract Foetal ECG from normal abdominal ECG

A deep neural network (DNN) was used by Iranian researchers to extract the foetal electrocardiogram (ECG) from a single abdominal ECG channel. Their process, which was published in the journal Physiological Measurement, may help with foetal monitoring in the future.

Currently, an ECG obtained from electrode patches mounted on the expectant mother's abdomen is used to measure the electrical activity of a foetus' heart. The foetal ECG can be used by clinicians to assess foetal health and diagnose anomalies.

Since the latter includes signals from both the foetus ("foetal ECG") and the mother ("maternal ECG"), as well as sources of interference including muscle contractions, it's difficult to distinguish the foetal ECG signal from abdominal ECG. When the amplitude of the foetal ECG signal is equal to that of the maternal ECG signal, this role becomes much more difficult.

Arash Rasti-Meymandi, the study's lead author and a graduate student at Iran University of Science and Technology, and his colleagues came up with a DNN-based solution to this issue.

Unets, convolutional networks commonly used in medical image segmentation tasks, inspired Rasti-Meymandi. To extract the maternal and foetal ECG signals, he and co-author Aboozar Ghaffari used a modified version of an Unet.

“The Unet was able to outperform other techniques in image segmentation challenges,” Rasti-Meymandi says. “To extract different components of the abdominal ECG, we examined the abdominal ECG signal at different resolutions, [similar to the process used in a Unet model].”

Using two sub-networks in sequence, the researchers' DNN, dubbed AECG-DecompNet, extracts the foetal ECG from a single channel of abdominal ECG. The maternal ECG signal is extracted by the first sub-network, while the foetal ECG is extracted by the second. The researchers used simulated ECG signals to train the two sub-networks separately, then tested them using both simulated and actual abdominal ECG recordings.

The researchers' DNN could process four seconds of abdominal ECG recording in about one second using a graphics processing unit.

DNNs and foetal ECG in the Future

Unlike other signal denoising methods that include the morphology (P, Q, R, S, and T waveforms indicating the heart's electrical activity) of a reference ECG, the researchers' method only requires a single channel of ECG. This not only makes ECG acquisition more comfortable for the mother, but it also takes less resources and time to implement than conventional ECG recording and signal extraction methods.

The researchers also discovered that their system preserved the shape and structure of foetal ECG signals better than other methods, with all five waveforms being well preserved for analysis and diagnosis of foetal abnormalities.

“The primary result of this research is the effectiveness of using DNNs to extract foetal ECG signal from a single-channel abdominal recording with high precision,” Rasti-Meymandi tells Physics World. “We are currently working on a more sophisticated algorithm … to further increase the accuracy of the extracted heart rate.”

The team is also developing ways to use their DNN on smartphones in real time.

Their approach has some drawbacks, including a possible overreliance on the training dataset, particularly when the foetal ECG signal is small, and the possibility of errors propagating from the first sub-network to the second.


To help their work, Newsmusk allows writers to use primary sources. White papers, government data, initial reporting, and interviews with industry experts are only a few examples. Where relevant, we also cite original research from others respected publishers.

Source- Physics World


2 views0 comments
bottom of page