Prediction of Potential Embryos for Implantation in the IVF Treatment Using Graph Convolution Network

Abstract

Nowadays, life style changes of human beings have created multiple challenges in obtaining normal pregnancy and leading to infertility. Infertility is considered as reproductive system disorder which occurs to large human population in the world due to lifestyle changes and some medical disorder. On advancement of the technologies, it become feasible to attain pregnancy artificially especially through treatments like In Vitro Fertilization (IVF), Assisted Reproductive Technology (ART) and Intracytoplasmic Sperm Injection (ICSI). Among those treatments, In Vitro Fertilization (IVF) becomes more familiar across peoples. Despite of several advantage of the In Vitro Fertilization (IVF) treatment, there exist some challenges in success rate of identifying potential embryo for implantation. Thus , deep learning model is only solutions which can increase the success rate of embryo implantation. In this paper, a new deep learning technique represented as graph convolution network is designed and implemented to identify the potential embryo for implantation. Graph Convolution Network composed of multiple layers and it processes the clinical data on transforming it into graph format. On processing, graph structured clinical data, it is highly efficient in extracting features of egg and organizes those extracted feature in form of the feature map. Finally fully connected layer of the model obtains the feature map in order predict the potential embryo for implantation using softmax function.  Especially proposed model is capable of determining the potential embryo towards its implantation in ovary region of the uterine wall. Particular model increases success rate of the fertility. Experimental analysis of the model is performed using clinical data obtained from the Apollo Hospital in the python environment. Obtained data is partitioned into training data and test data for model training and model testing. Model testing is performed as cross fold validation of the test data using confusion matrix to obtain the parameter for accuracy computation. On performance analysis, it is proved that proposed model outperforms state of art approaches.