Predicting high-intensity focused ultrasound efficacy in adenomyosis treatment based on magnetic resonance (MR) radiomics and clinical-imaging features.

Abstract

Aims: To develop a model predicting high-intensity focused ultrasound (HIFU) efficacy in adenomyosis treatment using enhanced T1WI and T2WI-FS radiomics combined with clinical imaging features.

Materials and methods: The study included 137 adenomyosis patients treated with HIFU from September 2021 to December 2023. Based on nonperfused volume ratio (NPVR), participants were divided into two groups: NPVR < 50% (n=77) and NPVR ≥ 50% (n=60). Patients were randomly split into training and test sets (7:3 ratio). Radiomics features were extracted from enhanced T1WI and T2WI-FS sequences, while clinical imaging features were selected using univariate analysis and binary logistic regression. Logistic regression models were built for radiomics, clinical imaging, and combined data. Model performance was assessed using ROC curves, Delong's test, and calibration curves.

Results: AUCs for the radiomics, clinical-imaging, and combined models in the training set were 0.831, 0.664, and 0.845, respectively, and 0.829, 0.597, and 0.831 in the test set. The combined model outperformed the clinical-imaging model (training p=0.001, test p=0.01) and the radiomics model (training p=0.012, test p=0.032). However, no significant difference was found between the combined and radiomics models (p>0.05). Calibration curves and decision curve analysis confirmed the combined model's accuracy and clinical applicability.

Conclusion: A model incorporating clinical-imaging features with T1WI and T2WI-FS radiomics effectively predicts HIFU success in adenomyosis treatment, offering valuable guidance for clinical decision-making.