Bridging the Diagnostic Gap between Histopathologic and Hysteroscopic Chronic Endometritis with Deep Learning Models

Area of Science:

• Reproductive Medicine
• Pathology
• Artificial Intelligence in Medicine

Background:

• Chronic endometritis (CE) is a uterine mucosal inflammation marked by CD138(+) endometrial stromal plasmacytes (ESPCs).
• CE is frequently observed in infertile women with various contributing factors, including unexplained infertility, tubal issues, endometriosis, implantation failure, and recurrent pregnancy loss.
• Traditional CE diagnosis via endometrial biopsy is invasive and may not capture the full mucosal picture.

Purpose of the Study:

• To develop and validate AI-driven prediction tools for histopathologic chronic endometritis using hysteroscopic images.
• To establish novel computer-aided detection and diagnosis systems for CE based on hysteroscopic findings.
• To offer a less-invasive diagnostic alternative for infertile women with suspected chronic endometritis.

Main Methods:

• The ARCHIPELAGO study utilized archival hysteroscopic images to train deep learning models.
• Models were developed to predict histopathologic CE based on visual findings during hysteroscopy.
• The study focused on creating computer-aided diagnostic tools for CE detection.

Main Results:

• Deep learning models were constructed to predict histopathologic CE from hysteroscopic images.
• The study aimed to create effective prediction tools for CE diagnosis.
• The developed models represent novel computer-aided systems for CE detection.

Conclusions:

• Hysteroscopy combined with deep learning offers a promising, less-invasive approach for diagnosing chronic endometritis.
• The ARCHIPELAGO study's AI models have the potential to significantly benefit infertile women suffering from CE.
• This AI-driven approach may improve the diagnostic accuracy and accessibility for chronic endometritis.