1. Data Acquisition
The wearable patch consists of EMG electrodes which are represented as 2, 4, and 6 in a triangular configuration. The FECG electrodes are represented as 1, 3, and 5 in the patch for obtaining the FHR and fetal presentation from the maternal abdomen. The electrodes are adhesive and need to be placed properly for recording the required physiological data. The electrodes are placed near the navel to capture the signals. The electrical activities of the electrodes are captured and they are transferred to the analytic IoT platform for feature extraction.
2. Interfacing with Analytic H-Iot Platform
In the analytic IoT platform “ThingSpeak”, a new channel is created which collects the physiological data wirelessly and sends it to the cloud for analysis by providing personal cloud for each user to store measurements giving API for easy access. The electrical activities of the electrodes are captured and the EMG and FECG data are sunk with the ThingSpeak. The channel specifications include the author’s name, channel ID and the access provided for the channel. The channel status includes the time duration of the created data, latest updates, and the last entry is done. The channel view can be shared with the family members and medical services for knowing the status of the patient.
3. Pre-Processing of Signals
The EMG signal is digitized at 100 samples per second and is digitally filtered using different 4-pole digital Butterworth filter from 0.3Hz to 4Hz in the IoT analyzer. The FECG signal obtained has a wandering baseline which may be due to the maternal motion artifacts, any electrical noises or even breathing movements that contributes to lower frequencies. Hence to avoid these effects, the mean value of the signal is subtracted from the signal itself. The power line interference is removed by using a notch filter with 50 Hz frequency. A Butterworth band pass filter of 4-80Hz is used to remove the DC components.
4. Number of Uterine Contractions Estimation
The uterine EMG signals measure the action potential changes associated with the uterine contractions of the pregnant woman. The uterine contractions depend on the intensity of the EMG signal as well as the duration of the event that occurs. Uterine contraction is a vital sign of labor. Hence monitoring the number of contractions is an essential parameter.
The average number of contractions for each stage is estimated as below. During the first stage, i.e. at the early labor contraction period, 4 contractions per hour will occur. Further, during the active labor contraction period, first stage, and 15 contractions per hour will occur. Finally, at the first stage, i.e. transition contraction period, 24 contractions per hour will occur. As the labor progresses, the second stage leads to 12 contractions per hour and the third stage have 2 contractions per hour. Hence the contractions increase as the delivery approaches and then it gradually decreases when the baby is delivered.
The average number of times the EMG signal crosses a particular threshold is denoted as Level Crossing Rate reflects the uterine contractile activity. The EMG signals produce bursts of action potential spikes, which provide information during the pregnancy and labor. The estimation of the threshold levels is calculated by overlaying the labor EMG signals and the non-labor EMG signals. Hence at the onset of labor, continuous burst of EMG signals is found to occur. The threshold is fixed as 1500 which contributes to a strong uterine contraction.
The uterine contractions are determined by detection of the peaks of the EMG signal. The minimum peak height is specified based on the intensity of the EMG signal, which denotes the signal crossing the fixed threshold levels. The location of the peaks is found by finding the maxima and ties of the peaks. The peaks are detected for the time-series which exceed the minimum peak height which represents the strong uterine contractions. On analysis, the thresholds are fixed based on certain criteria to differentiate between the contractions for true labor and Braxton-Hicks accordingly. The Fig. 2 shows an example in which the Level crossing rate is determined for the uterine EMG time series data and the contractions are estimated as 3 contractions which are denoted in red circles. Hence the 3 contractions represent the first stage of early labor.
5. Fetal Heart Rate Estimation
The FHR will normally accelerate during a uterine contraction, and then gradually slows as the mother and baby recovers. If the FHR fails to recover adequately, medical attention needs to be provided. The normal heart rate of the fetus ranges from 120 (bpm) to 160 (bpm) . Hence the safety margin is fixed as: 100 (bpm) < FHR (bpm) < 160 (bpm). If the FHR deviates from the safety margin then the fetal is prone to high risk of labor.
FECG signal provides valuable information about the fetal heart growth, fetal maturity, and health condition which is obtained by FECG electrodes on the surface of the abdomen. Maternal electrocardiogram (MECG) is a dominant noise mixed in FECG  and the amplitude, the magnitude, and the strength of MECG are greater than that of the FECG. Moreover, the baseline drift, the power-line interference, the gestational age, position of the electrodes, skin impedance and random electrical noise caused by human movement, baseline drift due to poor contact of measurement electrode are some external noises that can affect the FECG separation. The Essential problem is the efficient suppression of maternal electrocardiogram since its amplitude exceeds the level of the required signal. As the FECG signals are non-stationary and non-linear in nature, the noise suppression has to be done from the FECG signals. Suppression of maternal peaks for proper fetal signal extraction is required without losing the useful information.
In the wearable patch, the fetal electrical activity of the FECG electrodes placed on the abdomen is captured and sent to the ThingSpeak IoT platform wirelessly. The process of the FECG extraction is done as below.
A Discrete Wavelet Transform (DWT) is used in the ThingSpeak analysis to decompose the abdominal ECG signal which gives adaptive size window with maximum time-frequency resolution. DWT uses shorter windows at high frequencies (HF) and longer windows at low frequencies (LF) by applying a high pass filter and low pass filter. Hence a Daubechies wavelet is used as it is similar to the shape of the heartbeat. Then the data rate is reduced by a down sampling process.
The output of this down sampled data provides the detail and the approximate coefficients which analyze the HF and LF components respectively. The maternal QRS complexes are found from the approximate values of the obtained signal and the exact positions of the complexes are analyzed . From the analyzed positions, the maternal template is created. This template is cross-correlated with the maternal signal obtained from the clinical data provided. The best match which is correlated with the maternal signal is selected. The FECG signal is obtained after the MECG subtraction of the best correlating template. Now the IoT analyzer detects the fetal R wave peaks (fetal ECG wave peak) and it is denoted as R . The fetal R wave peak detection is done to calculate the RR interval (distance between successive R) . The calculated RR intervals are used for estimating the FHR as shown in Fig. 3.
The decision handler in the H-IoT platform checks for the safety margin of the Fetal Heart Rate (FHR) as follows
- 100<FHR<160: FHR normal.
- FHR>160: FHR is too high.
- FHR<100: FHR is too Low.
If it lies within the normal range, then the baby is in good health condition otherwise, medical attention should be given at the right time.
6. Determination of Fetal Presentation
The typical ECG adult heart using electrodes placed on the chest gives the QRS complex. If the positions of the adult chest electrodes reversed, their positions are rotated through 180°, the ECG becomes an inverse copy of the previous ECG. The method of rotating the electrode positions is similar to the context of the fetal ECG, to the fetus rotating through 180° within the maternal uterus. The FECG will register the intermediate shapes when the fetus undergoes an angular rotation of 90° within the maternal uterus.
The presentation of the fetus depends on the abdominal FECG waveforms . There are four types of FECG characteristic waveforms, where each one corresponds to a fetal presentation as Cephalic, Breech, Shoulder dorsoanterior and Shoulder dorsoposterior. The four types of FECG characteristic templates are generated and updated in the ThingSpeak IoT platform. The captured abdominal FECG signal is fed into another channel. Each template is cross-correlated with the captured signal and the maximum correlation with the templates is determined. The best-correlated template represents the fetal presentation accordingly as in Fig. 4 and Fig. 5 shows that type 1 ECG complex represents the Cephalic (Head down) presentation.
The decision handler in the IoT analyzer fixes safety limits for the three parameters. For EMG signals, the limits are fixed to differentiate between the contractions for true labor and Braxton-Hicks accordingly. For FECG signals, FHR normal range is fixed at 100 (bpm) < FHR (bpm) < 160 (bpm). For the fetal presentation, the signal template matching is done and presentations are found based on the highly correlated results.