Related Experiment Video
Updated: May 8, 2025
10:21
A Coregistered Ultrasound and Photoacoustic Imaging Protocol for the Transvaginal Imaging of Ovarian Lesions
Published on: March 3, 2023
1.3K
Integrative deep learning and radiomics analysis for ovarian tumor classification and diagnosis: a multicenter
Yi Zhou1, Yayang Duan1, Qiwei Zhu1
1Department of Ultrasound, The First Affiliated Hospital of Anhui Medical University, Shushan District, NO.218 Jixi Road, Hefei, 230022, Anhui Province, China.
La Radiologia Medica
|April 1, 2025
Summary
A combined deep learning and radiomics model using transvaginal ultrasound images accurately differentiates ovarian tumors. This advanced diagnostic tool improves malignancy prediction for better clinical decisions.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Oncology
Background:
- Ovarian tumors require accurate differentiation between benign and malignant types for effective treatment.
- Current diagnostic methods can have limitations in precision.
Purpose of the Study:
- To evaluate the combined effectiveness of transvaginal ultrasound (US)-based radiomics and a deep learning model for ovarian tumor classification.
- To assess the diagnostic performance against established methods like O-RADS and expert assessment.
Main Methods:
- A multicenter retrospective study analyzed grayscale and color US images from 2078 patients.
- Convolutional neural networks (CNNs) and radiomics models were developed and validated.
- A combined CNN-radiomics model was constructed and its predictive performance evaluated using AUC, sensitivity, and specificity.
Main Results:
- The combined CNN-radiomics model achieved the highest AUC (0.977 internal, 0.972 external).
- This model outperformed individual CNN and radiomics models, O-RADS, and expert assessments (p < 0.05).
- High sensitivity and specificity were observed for the combined model.
Conclusions:
- The combined CNN-radiomics model offers a highly accurate and reliable method for diagnosing ovarian tumors.
- This approach enhances malignancy prediction, providing clinicians with a more precise diagnostic tool.
- Transvaginal US-based radiomics and deep learning integration shows significant promise in gynecologic oncology.
