A Clinician-Friendly Machine Learning System to Predict Ovarian Response and Deploy Individualized Ovarian Stimulation Strategies in IVF

Reproductive health significantly influences both the overall health of individuals (e.g., cardiometabolic, mental and offspring congenital disorders) and human society. Ovarian stimulation (OS), the foundation of successful in vitro fertilization (IVF) treatments, has been complicated by the uncontrollability of ovarian response since OS was invented. This is mainly due to the unpredictable individual variability, long-term and complex therapies, a vast number of choices and limited evidence-based approaches for subgroups of responders. We developed a clinician-friendly machine learning (ML) based decision support system which can assist clinicians in: (1) diagnosing abnormal ovarian response earlier and faster, (2) understanding the pathogenic profiles of risk factors both in a population and at individual level, and (3) deploying individualized OS strategies that potentially strike a balance between the therapeutic effect and economic costs. It showed significant superiority to current ovarian reserve tests and their combinations that are widely used in clinical practice, and achieved excellent performance both in the internal and external validation. This ML system is also explainable, inspirational and still exhibits robustness in a number of stringent situations.

Funding Information: This research was supported by the National Key Research and Development Program of China (Grant No. 2018YFC1003201), the Natural Science Foundation of China (Grant nos. 82071604, 81803245), the Fundamental Research Funds for Central Universities (Grant No. 2021FZZX005-23), and the Natural Science Foundation of Zhejiang Province (Grant No. LQ15H04003).

Conflict of Interests: The authors declare no competing interests.

Ethical Approval: This study was approved by all local institutional review boards (the main center is the Zhejiang University School of Medicine Women’s Hospital, Approval Number: IRB-20200269-R. This study was also registered and approved in the Ningbo Women & Children Hospital, following the main center), waiving the need for informed consent due to the deidentified data.

Keywords: in vitro fertilization, poor ovarian response, hyper ovarian response, individualized ovarian stimulation, machine learning

Suggested Citation: Suggested Citation

Wang, Guiquan and Zhou, Liming and Li, Fengcheng and Hu, Xiaoling and Yu, Yun-Xian and Ying, Haochao and Chew, Ian and Zhu, Yimin, A Clinician-Friendly Machine Learning System to Predict Ovarian Response and Deploy Individualized Ovarian Stimulation Strategies in IVF. Available at SSRN: https://ssrn.com/abstract=4156167 or http://dx.doi.org/10.2139/ssrn.4156167

This version of the paper has not been formally peer reviewed.