Sengul, Y.A.Pacci, Z.Alagoz, O.Attar, R.2025-10-152025-10-152025979833152375610.1109/WIBI64774.2025.111152442-s2.0-105016311070https://doi.org/10.1109/WIBI64774.2025.11115244https://hdl.handle.net/20.500.14517/8473In-Vitro Fertilization (IVF) treatment typically begins with Controlled Ovarian Stimulation (COS), a process designed to stimulate the ovaries to produce multiple mature oocytes within a single cycle. COS plays a pivotal role in IVF success by improving the embryo implantation and pregnancy outcomes; however, it can also lead to complications such as ovarian hyperstimulation syndrome (OHSS). An optimal COS protocol aims to balance maximizing the chances of developing viable embryos while minimizing the risk of OHSS. In this work, we present a Machine Learning (ML)- based clinical decision support for selecting the optimal COS protocol for individual patients. Our approach begins with the development of a supervised classification model to predict the number of retrieved oocytes in 529 treatment cycles where OHSS complications did not occur. Among six different classification methods evaluated, Support Vector Classifier (SVC) achieved the highest performance, with 83.9% accuracy and an AUC value of 0.91. For new patients, we then calculate the probability of obtaining a sufficient number of oocytes under various COS protocols. The proposed method identifies the protocol with the highest likelihood of maximizing oocyte yield while reducing potential OHSS complications. © 2025 Elsevier B.V., All rights reserved.eninfo:eu-repo/semantics/closedAccessClinical Decision Support SystemControlled Ovarian StimulationIVF TreatmentConference ObjectN/AN/A