Diagnosis and Classification of Fetal Health Based on CTG Data Using Machine Learning Techniques

Abstract

Cardiotocograms (CTGs) is a simple and inexpensive way for healthcare providers to monitor fetal health, allowing them to take step to lessen infant as well as mother died. The technology operates by emitting ultrasound pulses and monitoring the response, revealing information such as fetal heart rate (FHR), fetal movements, uterine contractions, and more. Knowing the state of fetal, doctors and patients can take necessary steps in a time. Machine learning can play a vital role in this field. In this paper, we classified the state of fetal including normal state, suspect state, pathological state on the fetal disease dataset using seven machine learning model named AdaBoost (AdB), Random Forest (RF), K- nearest Neighbors (K-NN), Support Vector Machine (SVM), Gradient Boosting Classifier (GBC), Decision Tree Classifier (DTC), and Logistic Regression (LR). To validate the experimental task, we used several performance metrics containing accuracy, precision, recall, and F1-score. We also used a scaling technique named standard scalar for doing an unbiased dataset. Among the classification models, GCB outperforms the best by achieving the accuracy $95\%$, precision (for normal $96\%$, suspect $85\%$, pathological $97\%$), recall (for normal $98\%$, suspect $78\%$, pathological $94\%$), and F1-score (for normal $97\%$, suspect $81\%$, pathological $96\%$). Although, RF, SVM, and K-NN perform better precision ($100\%$) in the class of pathological state only.