Chemical Science & Engineering Research
Title
Performance Analysis of Various Supervised Classifiers for Predicting Preterm Delivery using Multi-channel Uterine EMG Signals
Authors
Suma K.V.,a Mamtha M.,*b and Poojac
abcDepartment of Electronics and Communication, M S Ramaiah Institute of Technology, Bangalore.
*Corresponding author E-mail address: mamtha.m@msrit.edu (Mamtha M.)
Article History
Publication details: Received: 12th July 2021; Revised: 18th October 2021; Accepted: 19th October 2021; Published: 2nd November 2021
Cite this article
Suma K.V.; Mamtha M.; Pooja. Performance Analysis of Various Supervised Classifiers for Predicting Preterm Delivery using Multi-channel Uterine EMG Signals. Chem. Sci. Eng. Res., 2021, 3(8), 7-11.
Abstract
Prediction of premature labour is of great significance in preventing infant deaths, or the consequent health risks globally. The enormous global burden on both families and society calls for preventive and predictive measures. The uterine Electromyography signals also called as Electrohysterogram (EHG) signals, has been very promising in studying the uterine contractions. Therefore, use of uterine EMG signals can prove to be a marker in diagnosing Preterm birth. In this study, the TPEHG DB (Term-Preterm Electrohysterogram Database) dataset with 300 records (262 term and 38 preterm records) are used. The raw uterine EMG signal is initiallypre-processedandthen various linear, non-linear and statistical features are extracted. The extracted features are applied to different machine learning classifiers. Further, Bayesian Hyper parameter Optimization technique was employed on these classifiers to improve their classification accuracy. Support vector machine (SVM) classifier with Bayesian Hyper parameter Optimization technique, tested using 10-fold cross-validation on 38 preterm records provided 96.667% accuracy.
Keywords
Uterine Electromyography; Premature labour; Bayesian Hyper parameter Optimization; Support vector machine
