CLESQ-ZOA-GCRNN: A Hybrid Deep Learning Framework for Predicting IVF/ICSI Success using Optimized Clinical Features

Important assisted reproductive technologies (ART) for treating infertility include in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI). Because of the wide variety of patient characteristics, treatment plans, and physiological reactions, their success rates are still low. This research presents a new hybrid model for IVF/ICSI treatment outcome prediction that combines a Graph Convolutional Recurrent Neural Network (GCRNN) with an improved metaheuristic optimization procedure called CLESQ-ZOA. The model makes use of a large fertility dataset that covers the years 2010-2018. It contains more than 490,000 treatment records that contain 94 clinical characteristics. A logarithmic spiral technique for improved exploration, chaotic mapping for diversified population initialization, and an increased solution quality mechanism for maintaining exploitation-exploration balance are the three primary advancements introduced by CLESQ-ZOA. By removing unnecessary or redundant inputs, these improvements allow for robust feature selection with precision. Next, the GCRNN is used to process the chosen feature subset. It integrates graph convolutional layers for spatial learning and LSTM units for temporal modeling. This two-pronged approach to modeling incorporates sequential patient data with inter-feature interactions. The suggested classical architecture outperforms baseline deep learning architectures and typical machine learning classifiers experimentally in terms of score and resilience. by forecasting the chance of successful IVF/ICSI outcomes, the approach gives actionable data for doctors, enabling tailored treatment planning. By enhancing patient results and alleviating the financial, emotional, and physical strain of multiple treatment cycles, this hybrid classical represents an encouraging advance in intelligent fertility diagnostics.