MODIFICATION OF PAN-TOMPKINS ALGORITHM FOR FETAL HEART RATE CALCULATION FROM THE MATERNAL ABDOMINAL ECG DATA: SLIDING WINDOW AND SQUARING SIGNAL RECONSTRUCTION

• Main Author: MIRZA ADHITAMA (NIM: 23214341), DONNY
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/26757
• Institution: Institut Teknologi Bandung

<p align="justify">Non-invasive fetal Heart Rate (fHR) monitoring technique become important for monitoring both fetal heart condition and stability in the gestational age. The fHR measurement commonly use an external CTG (Cardiotocograph) equipped with ultrasound doppler. The fHR is recorded on the CTG strip paper which usually less accurate because of various error. A new fHR monitoring technique based on abdominal Electrocardiogram (AECG) is developed in recent years. The fHR measurement by this new technique is still not satisfactory caused by difficulties to extract some parameters from the recorded signal. These parameters include the component of maternal QRS (mQRS) and fetal QRS (fQRS) complex with their different signal characteristics. The electrode placement on the mother’s abdomen also not quite standardized. In addition, fetal position inside the womb changes overtime. These conditions add more challenge for the researchers to extract fR-peak components which have much smaller amplitude than mR-peak components. <br /> <br /> <br /> Physionet published four-channel records of AECG signals for 2013 event Challenge. The dataset included the AECG signals (four-channels for each sample) and the fR reference point annotations for the comparison when performed the fHR estimation. In this study, part of the signals in this dataset were then being used to train the new fHR calculation algorithm while others then being used to test the new algorithm. <br /> <br /> <br /> The proposed method first passed through the signal filtering process by using a bandpass filter of 5-50Hz and 10-50Hz frequencies. Then, next process removed the mQRS complex components in the signal by using modified Pan-Tompkins algorithm. This algorithm also can be applied at the resulting of residual signal to detect the location of fR points with some adjustment. The resulting of the location of fR detected will be interpolated with some adjustment steps so the fR location detected could be more precise correspond to the fR reference location. If the input signal quality is poor, then some adjustments were made in filtering parameters to improve the fHR result. The adjustment included changing the value of the high frequency cutoff in the bandpass filter and also the threshold value at the sliding window creation int the mQRS complex detection stage. <br /> <br /> <br /> The result showed that the accuracy of the new developed fHR calculation algorithm was 92.4% for both system that applied BPF 5-50Hz and 10-50Hz. Physionet provided the scoring system to rate the fHR algorithm. The score for fHR algorithm developed in this study was 284. Although the acquired accuracy is quite good, the fHR algorithm developed in this study only rank number 20 compared to other developed algorithm published by Physionet In addition, the new developed fHR algorithm reduced the computational time by 45.9% compared to the original Pan-Tompkins algorithm.<p align="justify"> <br />