Extraction Of Features Of Uterine Emg Signals And Their Correlation

Main Article Content

Suma K V , Mamtha Mohan , Pooja

Abstract

Prediction of premature labor is of great significance in prevention of deaths of infants, or the risk of health ensuing. The uterine Electromyography signals have been very encouraging in the study of uterine contractions. Here, we have considered TPEHG DB (Term-Preterm Electrohysterogram Database) dataset having 300 records, of which 262 are term records while 38 are preterm records. Initially the raw uterine EMG signal is pre-processed and then various statistical,non-linear and linearfeatures are extracted. The features extracted are applied to different machine learning classifiers. Further, Bayesian Hyperparameter Optimization technique was employed on these classifiers to improve their classification accuracy. Support vector machine (SVM) classifier with Bayesian Hyperparameter Optimization technique was tested by employing a 10-fold cross-validation. This was conducted on 38 preterm records and it deliveredaccuracy of 96.667%. This is useful in early detection of preterm delivery in pregnant women and helps in avoiding infant fatality.

Article Details

Section
Articles