Mutual Information Grey Wolf Optimized and Gradient Derivative Recurrent Network Pregnancy Growth Data Analysis

Abstract

Introduction: Cardiotocography (CTG) is a clinical procedure that is utilized in tracing and measuring the extremity of fetal discomfort. Fetal health involves prediction of fetus health condition during pregnancy. Despite CTG is being the most frequently utilized equipment to keep track of and gauge the fetus health, the high rate of false positive results because of visual analysis appreciably contributes to unnecessary surgical delivery or delayed arbitration. By employing sophisticated techniques like, Artificial Intelligence (AI) and Deep Learning (DL) can boost the fetal health classification accuracy. Nevertheless, hardly any studies have concentrated on analyzing mutual dependence between features and non-linear transformation to the generated features making it potential in learning and performing more complicated tasks (i.e. predictive data analysis)
Objectives: To enhance maternal-fetal health and assist clinical decision-making by creating a more reliable and accurate AI-based fetal health classification system.
Methods: Mutual Information Grey Wolf Optimized Gradient Derivative Recurrent Network (MIGWO-GDRN) for pregnancy growth data analysis is proposed. The MIGWO-GDRN method is split into two parts, namely, feature selection and classification. First, the samples obtained from the Fetal Health Classification dataset as input are subjected to the Mutual Information-based Grey Wolf Optimization-based (MIGWO) feature selection model. The MIGWO optimizes the parameters by reflecting mutual dependence between two features aids in improving Pregnancy Growth Data Analytics accuracy. Second with the optimal selected features and the chief controlling features for classification being fetal heart rate, fetal movement and uterine contractions a DL method called, Backpropagated Gradient Derivative-based Recurrent Neural Network model (BPDRNN) is designed. The Backpropagated Gradient here allows to measure derivatives by means of three unique features: uterine contractions, fetal heart rate, and fetal movement respectively.
Results: From analysis, it is inferred that the MIGWO-GDRN method acquired an improved precision, recall and accuracy with minimal false positive and training time. This shows the improved performance of proposed pregnancy growth data analytic method
Conclusions: MIGWO-GDRN is presented for analyzing pregnancy growth data in order to improve the fetal health categorization procedure. The total performance results show that the provided MIGWO-GDRN technique achieves higher precision, recall, and accuracy with less training time and false positive rate when compared to existing methods.