Evaluation of Ultrasonic Doppler Signal Quality Based on Deep Learning

Fetal heart rate (FHR) monitoring is a crucial method for assessing a developing fetus’s well-being in the womb. It measures how many times the fetal heart beats per minute, offering insights into the baby’s vitality and whether there might be oxygen deficiencies in the uterus. Continuous FHR actions ensure the fetus’s safety. Accurate FHR interpretation can reduce unnecessary cesarean sections and lower the risk of fetal acidosis. The most widely used approach for FHR monitoring is the Doppler Ultrasound signal method. Nevertheless, a significant challenge in this method is the potential for signal loss due to interference and noise, which can lead to decreased signal quality and potentially lead to incorrect clinical assessments. Therefore, optimizing the monitoring of Doppler signal quality is essential to help obstetric caregivers make more precise fetal assessments, reduce unnecessary interventions, and enhance childbirth safety. Moreover, real-time feedback on the quality of Doppler Ultrasound signals can assist operators in making immediate adjustments to testing methods to improve signal quality, resulting in more dependable fetal heart rate data. This paper introduces an innovative deep learning network designed to classify the quality of Doppler Ultrasound signals. The model simplifies the complex data preprocessing for Doppler Ultrasound signals and integrates features from different levels to extract more essential and critical information, ultimately improving classification accuracy. As a result, the proposed system significantly enhances four key performance metrics: Accuracy, Precision, recall, and F1-score, surpassing the performance of existing methods. This improvement is aimed at increasing the accuracy and practical value of computer-assisted analysis systems, aiding obstetric healthcare professionals in result interpretation and decision-making.