A novel method for predicting pregnancy outcomes in IVF treatment: Deep Inception-ResNet

Infertility is a profound source of socioeconomic stress among numerous couples. Despite remarkable technological advancements in the field of assisted reproduction, couples still struggle for a clear understanding of their true chances of success. This issue has far-reaching consequences, both socially and medically, with live birth rates around only 25%. However, predictive models that could transform the landscape of IVF have given some hope.

In a groundbreaking study conducted by Majumdar G. et al. at the Center of IVF and Human Reproduction, Sir Ganga Ram Hospital, clinical data from 2268 patients who underwent IVF/ICSI procedures between January 2018 and December 2020 were meticulously analyzed. The predictive model considered maternal age and the number of IVF cycles to the type and duration of infertility, anti-Müllerian hormone (AMH) levels, indication for IVF, sperm type, body mass index (BMI), embryo transfer specifics, and β-human chorionic gonadotropin (β-hCG) values.

When tested with an 80:20 train-test split and feature selection, the model demonstrated an accuracy rate of 76% and a ROC-AUC score of 0.80. Notably, this tool is a novel happening in reproductive health research, which can guide infertile couples to make more informed decisions before starting an IVF. 

In conclusion, this study marks a significant step forward in improving IVF outcomes and reducing the emotional and financial toll of infertility. 

Majumdar G, Sengupta A, Narad P. et al. Deep Inception-ResNet: A Novel Approach for Personalized Prediction of Cumulative Pregnancy Outcomes in Vitro Fertilization Treatment (IVF). J ObstetGynecol India. 2023. https://doi.org/10.1007/s13224-023-01773-9