An AI-Driven Clinical Decision Support Model Based on Anemia and Fibroid Parameters to Guide Surgical Decision-Making

Area of Science:

• Gynecology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Uterine fibroids (UFs) are common, and surgical intervention decisions can be subjective.
• Identifying objective factors for myomectomy is crucial for patient care.

Purpose of the Study:

• To identify clinical factors predicting the need for surgical intervention in women with UFs.
• To develop and validate a data-driven clinical decision helper algorithm for myomectomy.
• To assess the algorithm's concordance with expert gynecological assessment.

Main Methods:

• Retrospective analysis of 618 women with UFs across three hospitals.
• Comparative statistical analysis between surgical (myomectomy) and non-surgical groups.
• Development and validation of machine learning (ML) models using fibroid characteristics and anemia markers (hemoglobin, ferritin).
• Prospective clinical concordance assessment with 50 real-time cases evaluated by a blinded gynecologist.

Main Results:

• The surgery group showed significantly lower hemoglobin and ferritin levels (p < 0.001).
• ML models, particularly support vector machines integrating ferritin levels with UF characteristics, achieved 98-99% accuracy.
• Ferritin-based models demonstrated high precision, sensitivity, and F1-scores.
• The algorithm achieved 98% agreement with blinded expert gynecologist assessments.

Conclusions:

• A highly accurate, objective decision helper algorithm for predicting surgical necessity in UF patients was developed.
• Anemia status (ferritin) and fibroid characteristics are key predictors for myomectomy.
• The algorithm reduces subjective variability and supports integration into routine gynecologic decision-making.