An Intelligent Fetal Prognostic System using Machine Learning Techniques

Pregnancy, an intricate journey emblematic of women’s resilience, necessitates comprehensive prenatal care for the well-being of both mother and fetus. This research proposes an innovative approach by integrating cardiotocography data and ultrasound scans to enhance fetal health predictions. Genetic algorithms are employed to identify the most relevant features in the cardiotocogram dataset, with a subsequent comparison of various algorithms to select the one producing optimal evaluation metrics. Using genetic algorithms allows a reduction in training time and building of a more robust model. The extracted features are then utilized to train diverse machine learning models, including Artificial Neural Networks, K-Nearest Neighbours, Support Vector Machine, and Random Forest. The resulting machine learning model demonstrates the capability to predict crucial fetal parameters such as morphological patterns and the fetal state. In the context of measuring head circumference from ultrasound images, convolutional neural networks was used to perform segmentation and to extract the region of interest. To determine the head circumference, two equations were employed, i.e. Ramanujan approximation of ellipse circumference and a standard formula. Evaluation metrics such as dice coefficient, and Hausdorff distance are used to rigorously assess the integrated system’s performance. This multifaceted approach strives to revolutionize prenatal care by leveraging machine learning techniques.