Artificial Intelligence, Deep Learning, and Computer Vision in Hysteroscopy: A Systematic Review

Area of Science:

• Gynecological Endoscopy
• Medical Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Hysteroscopy is the standard for diagnosing and treating intrauterine conditions.
• Interpretation of hysteroscopy is operator-dependent, leading to variability.
• Artificial intelligence (AI), machine learning (ML), deep learning (DL), and computer-aided diagnosis (CAD) offer potential for improved consistency and decision support.

Purpose of the Study:

• To systematically review AI, ML, DL, and CAD applications in hysteroscopy.
• To assess the performance and limitations of these technologies in gynecological procedures.

Main Methods:

• Systematic literature search of PubMed/MEDLINE and EBSCOhost up to March 8, 2026.
• Inclusion of 19 primary studies across various AI applications in hysteroscopy.
• Risk of bias and technical quality assessment using QUADAS-2, PROBAST, and RoB2.

Main Results:

• AI demonstrated high performance in specific tasks like binary classification (AUC up to 0.979) and fertility prediction (AUC up to 0.992).
• Applications included diagnostic classification, lesion detection, segmentation, and prognostic support.
• Most studies were retrospective, single-center, and lacked external validation; only one randomized study linked AI to outcomes.

Conclusions:

• AI shows technical proficiency in selected hysteroscopic tasks, including binary classification and lesion detection.
• Current evidence is constrained by study design limitations, including retrospective data and operator-dependent image acquisition.
• Near-term applications focus on supporting image interpretation, quality control, lesion highlighting, and data integration.