Heart rate estimation and validation algorithm for fetal phonocardiography

Objective. Fetal heart rate (FHR) is an important parameter for assessing fetal well-being and is usually measured by doppler ultrasound. Fetal phonocardiography can provide non-invasive, easy-to-use and passive alternative for fetal monitoring method if reliable FHR measurements can be made even in noisy clinical environments. Therefore, this work presents an automatic algorithm to determine FHR from the fetal heart sound recordings in a noisy clinical environment. Approach. Using an electronic stethoscope fetal heart sounds were recorded from the expecting mother's abdomen, during weeks 30-40 of their pregnancy. Of these, 60 recordings were analyzed manually by two observers to obtain reference heart rate measurement. An algorithm was developed to determine FHR using envelope detection and autocorrelation of the signals. Algorithm performance was improved by implementing peak validation algorithm utilizing knowledge of valid FHR from prior windows and power spectral density function. The improvements in accuracy and reliability of algorithm were measured by mean absolute error (MAE) and positive percent agreement. Main results. By including the validation step, the MAE reduced from 11.50 to 7.54 beats per minute and positive percent agreement improved from 81% to 87%. Significance. We classified the recordings into good, moderate and poor quality to understand how the algorithm works in each of the case. The proposed algorithms provide good accuracy overall but are sensitive to the noises in recording environment that influence the quality of the signals.