The role of shear wave elastography in predicting clinical symptoms in adenomyosis: A prospective observational study with a machine learning approach

OBJECTIVE: Adenomyosis is a gynecological condition characterized by the invasion of endometrial tissue into the myometrium, causing symptoms such as dysmenorrhea, menorrhagia, and chronic pelvic pain. Its diagnosis remains challenging due to overlapping features with other uterine disorders, and the variability in symptom presentation makes management complex. This study aims to evaluate the utility of shear wave elastography (SWE) in predicting clinical symptoms of adenomyosis and to explore the potential of machine learning (ML) models in enhancing diagnostic precision and predicting patient outcomes. METHODS: A total of 63 women diagnosed with focal adenomyosis were included in this prospective observational study. SWE was performed to assess uterine tissue stiffness, with ML algorithms (logistic regression, random forest, support vector machine, K-nearest neighbors) applied to predict symptoms based on SWE measurements and clinical features. Clinical symptoms such as dysmenorrhea, dyspareunia, non-cyclic chronic pelvic pain, and menorrhagia were evaluated. Statistical analysis was conducted using SPSS software, with performance metrics such as accuracy, F1 score, and ROC-AUC used to assess model effectiveness. RESULTS: Significant associations were found between SWE velocity (SWV) values and symptoms like dysmenorrhea, dyspareunia, and non-cyclic chronic pelvic pain. K-nearest neighbors (KNN) exhibited the best performance in predicting dyspareunia and non-cyclic chronic pelvic pain, while random forest performed best for dysmenorrhea. Menorrhagia did not show significant differences in SWE values. Cutoff values for clinical symptoms, such as 4.69 m/s for dysmenorrhea, were identified, providing actionable thresholds for clinical use. CONCLUSION: SWE combined with ML offers a promising approach to predict clinical symptoms of adenomyosis, aiding in personalized treatment strategies. This study highlights the potential of integrating advanced imaging techniques and computational models to enhance clinical decision-making and improve patient outcomes.