Multi-modality imaging technologies and machine learning for non-invasive, precise assessment of rabbit endometrium

Area of Science:

• Reproductive Biology
• Biomedical Optics
• Medical Imaging

Background:

• Accurate assessment of endometrial conditions is crucial for improving pregnancy rates in assisted reproductive technology (ART).
• Current diagnostic methods often lack the real-time, minimally invasive capabilities needed for dynamic endometrial monitoring.

Purpose of the Study:

• To develop and validate a minimally invasive, real-time diagnostic tool for assessing endometrial status.
• To enhance the accuracy of endometrial receptivity assessment using advanced imaging and machine learning.

Main Methods:

• Utilized fiberoptic bronchoscopy and optical coherence tomography (OCT) to monitor rabbit endometrium changes.
• Performed histological analysis and electron microscopy alongside training a conditional generative adversarial network (cGAN).
• Developed a cGAN to convert OCT images into virtual hematoxylin and eosin (H&E) stained sections.

Main Results:

• Successfully non-invasively observed dynamic changes in the endometrium at various stages.
• Demonstrated OCT's capability to provide detailed superficial and submucosal views, correlating with H&E pathology.
• Showcased machine learning's ability to enhance OCT by generating H&E-like images.

Conclusions:

• The combined optical imaging and machine learning approach enables real-time, non-invasive endometrial assessment.
• This technology has the potential to significantly improve the accuracy of endometrial receptivity evaluation in ART.
• The developed tool offers a promising alternative to traditional methods for monitoring endometrial health.