MRI-Based Radiomics as a Promising Noninvasive Diagnostic Technique for Adenomyosis

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Magnetic Resonance Imaging (MRI) is crucial for diagnosing adenomyosis, particularly when ultrasound results are inconclusive.
• Adenomyosis diagnosis can be challenging, necessitating advanced imaging techniques.

Purpose of the Study:

• To evaluate the efficacy of MRI-based radiomics in distinguishing uteri with and without adenomyosis.
• To explore radiomics as a potential imaging biomarker for adenomyosis.

Main Methods:

• A retrospective single-center study involving 15 patients (9 with adenomyosis, 6 controls).
• Preoperative T2w MRI scans were analyzed using 3D segmentation and extraction of 884 radiomics features.
• Machine learning models were developed after dimension reduction and feature selection, with diagnostic performance assessed via ROC analysis.

Main Results:

• Eleven significant radiomics features were identified for differentiating adenomyosis.
• Individual features demonstrated high diagnostic performance (AUCs 0.78–0.98).
• Machine learning models incorporating multiple features achieved excellent performance (AUC 1.0).

Conclusions:

• Radiomics features from routine T2w MRI accurately differentiate adenomyosis.
• Radiomics holds promise for enhancing adenomyosis diagnosis and serving as an imaging biomarker for therapy personalization and treatment monitoring.