A novel fetal ECG signal compression method: Variance and neighboring based data compression

Graphs of ECG signals in time domain have an important place in the diagnosis of heart diseases. ECG (fetal ECG-FECG) signals from fetuses also reveal heart problems, particularly changes in ST interval, that the baby experiences before birth. The rapid, reliable and cost-effective transmission of health data through telemedicine applications is increasing when the distance between health service providers and health care providers increases. In this work, a new method for compressing FECG signals received from the PhysioNet database is proposed. This proposed method is variance and neighborhood based data compression (VNBDC) algorithm. In the signal compression, the FECG signals were divided into 5 ms, 10 ms, and 20 ms segments. Then, the column-based variances of each segmented FECG signal were calculated. Differences between binary adjacent variance values were calculated through the column and then a threshold value was determined by averaging these differences between them. In the FECG signal, the variances of adjacent distributions on a column basis were compared with this threshold value, and those larger than this threshold value were discarded and the signal was compressed by removing similar columns. In the study, a non-interference fetal ECG database was used and it was observed that there was not a significant change (mean change 0.05%) in the time domain features of FECG signals compressed with raw FECG signal, while a compression of approximately 20% to 40% was obtained in the end result. Based on the results obtained, this proposed compression method has been shown to be safely used in biotelemetry systems.