Automated assessment of endometrial receptivity for screening recurrent pregnancy loss risk using deep learning-enhanced ultrasound and clinical data
2024

Using Deep Learning to Assess Endometrial Receptivity for Recurrent Pregnancy Loss Risk

Sample size: 715 publication 10 minutes Evidence: high

Author Information

Author(s): Yan Shanling, Xiong Fei, Xin Yanfen, Zhou Zhuyu, Liu Wanqing

Primary Institution: Deyang People’s Hospital, Deyang, Sichuan, China

Hypothesis

Can deep learning techniques improve the assessment of endometrial receptivity in women with recurrent pregnancy loss?

Conclusion

The study shows that a deep learning-enhanced fusion model can accurately stratify the risk of recurrent pregnancy loss, outperforming traditional methods.

Supporting Evidence

  • The ResNet-50 model achieved the highest accuracy and lowest Brier score among deep learning architectures.
  • The fusion model provided the most accurate predictions with an area under the curve of 0.853.
  • Deep learning models outperformed traditional machine learning models in predicting recurrent pregnancy loss risk.

Takeaway

This study used smart computer programs to look at ultrasound images and health data to help doctors figure out which women might have trouble getting pregnant again.

Methodology

The study used a retrospective, controlled design with deep learning models analyzing ultrasound images and clinical data from participants.

Potential Biases

Potential bias due to the retrospective design and single-center data collection.

Limitations

The sample size is relatively small and sourced from a single clinical center, which may affect generalizability.

Participant Demographics

Women aged 20-40 years with unexplained recurrent pregnancy loss and controls who achieved full-term pregnancies.

Statistical Information

P-Value

<0.001

Confidence Interval

1.157–1.310

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.3389/fphys.2024.1404418

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication