Extracorporeal life support

Abstract

Background

With the global increase of Cesarean section delivery rates, the long-term effects of Cesarean delivery have started to become more clear. One of the most prominent complications of Cesarean section in recurrent pregnancies is uterine rupture. Assessing the risk of uterine rupture or dehiscence is very important in order to prevent untimely operations and/or maternal and fetal complications.

Objective

Our study aims to assess whether machine learning can be used to predict uterine dehiscence or rupture by using patients' ultrasonographic findings, clinical findings and demographic data as features. Hence, possible uterine rupture, as well as maternal and fetal complications pertinent to it, could be prevented.

Study Design

The study was conducted on 317 patients with term (>37 weeks) singleton pregnancy. Demographics, body-mass indices, smoking and drinking habits, clinical features, past pregnancies, number and history of abortions, inter-delivery period, gestation week, number of previous Cesarean operations, fetal presentation, fetal weight, tocography data, trans-abdominal ultrasonographic measurement of lower uterine segment full thickness and myometrium thickness, lower uterine segment findings during Cesarean section were collected and analyzed using machine learning techniques. Logistic Regression, Multilayer Perceptron, Support Vector Machine, Random Forest and Naive Bayes algorithms were used for classification. The dataset was evaluated using 10-fold cross-validation. Correct Classification Rate, F-score, Matthews Correlation Coefficient, Precision-Recall Curve area and Receiver Operating Characteristics area were used as performance metrics.

Results

Among the machine learning techniques that has been tested in this study, Naive Bayes algorithm showed the best prediction performance. Among the various combinations of features used for prediction, the essential features of parity, gravida, tocographic contraction, dilation, d&c with the sonographic thickness of lower uterine segment myometrium yielded the best results. The runner-up performance was obtained with the sonographic full thickness of lower uterine segment added to the base features. The base features alone can classify patients with 90.5% accuracy, while adding the myometrium measurement increases the classification performance by 5.1% to 95.6%. Adding the full thickness measurement to the base features raises the classification performance by 4.8% to 95.3% in terms of Correct Classification Rate.

Conclusion

Naive Bayes algorithm can correctly classify uterine rupture or dehiscence with a Correct Classification Rate of 0.953, an F-score of 0.952 and a Matthews Correlation Coefficient value of 0.641. This result can be interpreted such that by using clinical features and lower uterine segment ultrasonography findings, machine learning can be used to accurately predict uterine rupture or dehiscence.

Trial registration

Clinical Research Ethics Committee of Ankara City Hospital, University of Health Sciences (Approval number: E2-20-108) Date of registration: 27-01-2021 URL: https://ankarasehir.saglik.gov.tr/TR-348810/2-nolu-etik-kurul.html e-mail: ankarash.etikkurul2@saglik.gov.tr

AJOG at a Glance

A.

Why was this study conducted?

This study was conducted to: Determine whether machine learning algorithms can be utilized to predict uterine dehiscence and assess the risk of uterine rupture Evaluate the contribution of ultrasonographic measurement of lower uterine segment measurements to the prediction performance of the algorithms Find out which machine learning technique performs the best for predicting uterine dehiscence. B.

What are the key findings?

Machine learning methods can be used to accurately predict uterine dehiscence (with up to 95.6% accuracy). Using lower uterine segment full thickness or myometrium thickness increases the accuracy of Naive Bayes algorithm by 4.8% and 5.1%, respectively. Naive Bayes algorithm yields the best prediction performance among the methods tried. C.

What does this study add to what is already known?

Ultrasonographic lower uterine segment measurements can be used as features in machine learning to increase its prediction performance of uterine dehiscence and hence the risk of uterine rupture.