Journal's Facebook pageApplication of the Inception-ResNet-V2 algorithm to the analysis of embryo microscope images for the prediction model of assisted reproduction
- Department of Biomedical Engineering
National Yang Ming Chiao Tung University, Taipei, Taiwan
sheepkelly19.be11@nycu.edu.tw (Y.-Y.Y.), wchu@nycu.edu.tw (W.-C.C.) - IHMED Reproductive Center
Taipei 106028, Taiwan
drweng@ihmed.com.tw - Department of Biomedical Science and Engineering
National Central University, Taoyuan 320317, Taiwan
sulijen@gmail.com - Department of Information Management
Shih Hsin University, Taipei, Taiwan
kjhtw@mail.shu.edu.tw
Abstract
The World Health Organization (WHO) estimates that approximately14 80 million men and women worldwide, with childbearing potential, need medical assistance due to fertility difficulties, which represents approximately 15% of the population. Similarly, about 15% of couples of maternal ages in Taiwan experience infertility problems. In clinical practice, in vitro fertilization (IVF) is the primary method of artificial reproduction. Using deep learning technology and an Incep-tion-ResNetV2 model, we can create a reliable embryo classification and prediction system, which improves the selection of high-quality embryos and enhances pregnancy success rates. The classification and prediction model achieved 80% precision, AUC= 0.88, sensitivity 73% and 88% specificity. This exceeds the statistics of the Taiwanese National Health Service, where the average pregnancy rate for IVF in 2023 was 27.8 %. The results indicate that our model efficiently classifies embryos for successful implantation at a higher rate than the national statistics in Taiwan.
Key words
Convolution neural networks, Deep learning; embryo, In vitro fertilization
Digital Object Identifier (DOI)
https://doi.org/10.2298/CSIS241001066Y
How to cite
Yen, Y., Weng, S., Su, L., Kao, J., Chu, W.: Application of the Inception-ResNet-V2 algorithm to the analysis of embryo microscope images for the prediction model of assisted reproduction. Computer Science and Information Systems, https://doi.org/10.2298/CSIS241001066Y
