Model Based on Ultrasound Radiomics and Machine Learning for Differentiating Uterine Fibroids and Adenomyomas.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Gynecologic Oncology

Background:

• Differentiating uterine fibroids (UFs) and adenomyomas (AMs) is crucial for effective patient management.
• Current diagnostic methods can be invasive or lack definitive accuracy.
• Developing noninvasive tools for accurate differentiation is a significant clinical need.

Purpose of the Study:

• To assess the diagnostic value of ultrasound radiomics in distinguishing between UFs and AMs.
• To develop and validate a noninvasive tool for differentiating these conditions.
• To improve preoperative planning and fertility preservation strategies.

Main Methods:

• Retrospective analysis of ultrasound images and clinical data from 659 patients with UFs or AMs across three institutions.
• Extraction of radiomics features from ultrasound images.
• Development of clinical, radiomics (SVM-based), and combined models using machine learning classifiers.
• Validation of model performance using AUC, calibration curves, and decision curves.

Main Results:

• The combined model achieved the highest diagnostic performance, with AUCs of 0.923 in the training cohort and 0.894 in the external test cohort.
• The SVM-based radiomics model demonstrated strong discrimination, with AUCs of 0.899 (training) and 0.823 (test).
• The combined model showed excellent calibration and the greatest clinical utility compared to individual models.

Conclusions:

• Ultrasound radiomics, particularly with an SVM approach, is a feasible method for differentiating UFs and AMs.
• The developed combined model serves as a reliable and effective noninvasive tool for differentiating UFs and AMs.
• This tool can significantly aid clinicians in preoperative decision-making and fertility preservation planning.