Ultrasonographic cervical length assessment is increasingly being utilized clinically to identify women at risk for spontaneous preterm delivery. In a randomised prospective longitudinal study involving 200 women, we measured cervical length and internal os diameter by transvaginal scan at 20 - 24 weeks and analysed their ability to predict preterm birth. The risk of spontaneous preterm delivery increased steeply as cervical length decreased. At cut off value of < or = 2.5cm, the cervical length measurements had sensitivity, specificity, positive predictive value and negative predictive value of 77%, 95%, 56% and 98% respectively. However, internal os diameter lacked sensitivity and specificity. Our data suggests that the duration of pregnancy is directly related to length of the cervix: the shorter the cervix, the greater the chance of preterm delivery.
An accurate detection of preterm labor and the risk of preterm delivery before 37 weeks of gestational age is crucial to increase the chance of survival rate for both mother and the infant. Thus, the uterine contractions measured using uterine electromyogram (EMG) or electro hysterogram (EHG) need to have high sensitivity in the detection of true preterm labor signs. However, visual observation and manual interpretation of EHG signals at the time of emergency situation may lead to errors. Therefore, the employment of computer-based approaches can assist in fast and accurate detection during the emergency situation. This work proposes a novel algorithm using empirical mode decomposition (EMD) combined with wavelet packet decomposition (WPD), for automated prediction of pregnant women going to have premature delivery by using uterine EMG signals. The EMD is performed up to 11 levels on the normal and preterm EHG signals to obtain the different intrinsic mode functions (IMFs). These IMFs are further subjected to 6 levels of WPD and from the obtained coefficients, eight different features are extracted. From these extracted features, only the significant features are selected using particle swarm optimization (PSO) method and selected features are ranked by Bhattacharyya technique. All the ranked features are fed to support vector machine (SVM) classifier for automated differentiation and achieved an accuracy of 96.25%, sensitivity of 95.08%, and specificity of 97.33% using only ten EHG signal features. Our proposed algorithm can be used in gynecology departments of hospitals to predict the preterm or normal delivery of pregnant women.