Detection of Follicles using Improved Salp Swarm Algorithm as a vital Tool in the Resolution of Infertility in Women

Abstract

Follicle is a sac filled with fluid in the woman’s reproductive organ and it usually appears darker compared to other objects in the ovary. The ovarian follicles and their sizes are imaged by ultrasound machine. The detection of these follicles is affected usually by other organs such as tissues, blood vessels contained in the ovary. Thus, the need to process the image using appropriate method in order to detect, and extract the required features for better classification of the follicles and non-follicles present in the ovarian image. Feature extraction is a major task in follicles classification of scanned image, therefore proper analysis of image features provides useful information that aid detection of follicle and non-follicle spot in the ovary. This work considers an Improved Salp Swarm Algorithm known as Elite Opposition Based Learning Salp Swarm Algorithm to select features with best fitness from the extracted features. The selected features were served as input to the Multilayer Perceptron Artificial Neural Network classifier. The objective is to evaluate the performance of the developed system based on the benchmark figures already obtained by other researchers in terms of follicle detection rate such as accuracy, sensitivity and specificity of the system. Upon training and testing the network in terms of the False Acceptance Ratio and False Rejection Ratio, 98.6% accuracy, 100% sensitivity and 98.1% specificity were achieved. These results give a better and more reliable classification output that will facilitate the diagnosis of infertility and other related reproductive diseases in women.