Title: Classification of Fetal Cariotocography / Xavier Robinston Antony.
Material Type: Book
Creator: Antony, Xavier R. 1969- author. (Xavier Robinston),
Subject: Fetal heart rate monitoring--Classification.
Description: Fetal Heart Rate (FHR) monitoring is the process of checking the condition of the fetus during labor and delivery by monitoring fetal heart rate with special equipment. Uterine Contractions (UC) generally signal that labor is starting. The abdomen becomes hard during contractions and the uterus relaxes and the abdomen becomes soft between contractions. Cardiotocography is a simultaneous recording of fetal heart rate (FHR) and uterine contractions (UC). Fetal heart rate monitoring may help detect changes in the normal heart rate pattern during labor and help prevent unnecessary treatments that are not needed. It is one of the common diagnostic techniques to evaluate mother and fetal well-being and facilitating physicians to take necessary actions during labor and post labor. The dataset consists of measurements of fetal heart rate (FHR) and uterine contraction (UC) features on cardiotocograms classified by expert obstetricians. 2126 fetal cardiotocograms (CTGs) were automatically processed and the respective diagnostic features measured. The CTGs were also classified by three expert obstetricians and a consensus classification label assigned to each of them. Therefore, the dataset can be used either for 10-class or 3-class experiments. This project thesis concentrates on 3-class such as Normal (N) state, Suspicious (S) state and Pathological (P) state classification Cardiotocogram data with 21 predictor variables and target variable. Exploratory Data Analysis (EDA) is performed to find the correlation between predictor variables and correlation between predictor variables and target variable. Histograms and normalized histograms of predictor variables are analyzed to understand the data distribution and relationship between predictor and target variables. Clustering analysis performed to seek to uncover groups of records with similar behavior. K-Means Clustering is applied on each predictor variable to uncover the groups of records with similar behaviors and the clustering result is overlaid with the target variable NSP to find the relation between cluster and target variable. Principal Component analysis seeks to uncover groups of variables with similar behavior. Multicollinearity is a condition where some of the predictor variables are correlated to each other. Such a condition is not good for building models. These variables can be combined together as another variable using Principal Component Analysis. Various criteria are analyzed to find the suitable number of principal components. In Classification Analysis, there are six models are built using with and without misclassification costs and model evaluation measures of all models are calculated. The Accuracies, Overall Error Rate and other evaluation measures are compared and the best model is selected. The misclassification cost applied C5.0 model using the original variables gives maximum accuracy and verified using testing dataset.
Date of Publication: 2015.
Publisher: Central Connecticut State University,
Date: 2015