Automatic segmentation of uterine contractions in EHG signals: Hardware implementation with Raspberry Pi and Arduino Mega

Abstract

Premature birth is one of the major problems in obstetrics. Early detection of preterm birth is an important key to prevent and reduce its consequences. As a result, it has been a subject of interest for many researchers. The current work is placed as part of the development of a non-invasive clinical aid tool. We implement a portable and easy-to-use device for pregnant women be able to automatically detect signs associated with uterine contractions. The device is able to detect EHG segments associated with uterine contractions during pregnancy. EHG signals are acquired, in real time, from 12 bipolar electrodes placed on the abdomen of pregnant women. The data are then analyzed and processed with a powerful processor. The technique used for detection is based on the Dynamic Cumulative Sum (DCS) method which has already demonstrated its strong potential in event detection when applied to non-stationary signals. The DCS method is followed by a data fusion method to merge the segments of all bipolar channels. A technique based on Fisher's test has been implemented and applied between two consecutive segments in order to reduce the over-segmentation problem. This strategy was proposed in previous studies and was proved to be accurate and efficient. The results show that a prototype based on the Raspberry Pi board and the Arduino Mega board is very promising for setting up a final prototype.