Automated detection of preterm condition using uterine electromyography based topological features

Accurate prediction of preterm birth is a global, public health priority. This necessitates the need for an efficient technique that aids in early diagnosis. The objective of this study is to develop an automated system for an effective detection of preterm (weeks of gestation < 37) condition using Electrohysterography (EHG) and topological features associated with the frequency components of signals. The EHG signals recorded prior to gestational age of 26 weeks are considered. The pre-processed signals are subjected to discrete Fourier transform to obtain the Fourier coefficients. The envelope is computed from the boundary of the complex Fourier coefficients identified using the alpha-shape method. Topological features namely, area, perimeter, circularity, convexity, ellipse variance and bending energy are extracted from the envelope. Classifications based on threshold-determination method and machine learning algorithms namely, naive Bayes, decision tree and random forest are employed to differentiate the term and preterm conditions. The results show that the Fourier coefficients of EHG signals exhibit different shapes in the term and preterm conditions. The regularity of signals is found to increase in preterm condition. All the features are found to have significant differences between these two conditions. Bending energy as a single biomarker achieves a maximum accuracy of 80.7%. The random forest model based on the topological features detects the conditions with the maximum accuracy and positive predictive value of about 98.6%. Therefore, the proposed automated system seems to be effective and could be used for the accurate detection of term and preterm conditions. (c) 2021 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences. Published by Elsevier B.V. All rights reserved.

Automated summary

This study aims to develop an automated system for effective detection of preterm conditions using EHG signals and topological features. Results showed that the system could accurately differentiate term and preterm conditions with the maximum accuracy and positive predictive value of about 98.6%.