Discussion on the accuracy of ovarian tumor diagnosis based on artificial intelligence with different scanning methods
10.13491/j.issn.1004-714X.2025.01.013
- VernacularTitle:基于人工智能诊断不同扫查方式的卵巢肿瘤准确性的讨论
- Author:
Haizheng WANG
1
;
Li FENG
1
;
Sen WANG
1
;
Huimin GUO
1
;
Fanguo MENG
1
Author Information
1. Department of Medical Ultrasound, The First Affiliated Hospital of Shandong First Medical University & Shandong Provincial Qian Foshan Hospital, Shandong Medicine and Health Key Laboratory of Abdominal Medical Imaging, Jinan 250014 China.
- Publication Type:OriginalArticles
- Keywords:
Ovarian cancer;
Transvaginal ultrasound;
Transabdominal ultrasound;
CNN;
Segmentation;
Ovarian cyst
- From:
Chinese Journal of Radiological Health
2025;34(1):77-83
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the accuracy of artificial intelligence-based diagnosis of ovarian malignant tumors and the identification of benign and malignant tumors under transabdominal scanning and transvaginal scanning methods. Methods A dataset of transabdominal and transvaginal two-dimensional ultrasound images was used and the images were preprocessed to enhance quality. The region of interest was segmented and divided into a training set and a test set. A convolutional neural network (CNN) was trained on the images in the training set, and the accuracy of the model on the test set was calculated. Results Transvaginal scanning was 14% more accurate in diagnosing malignant ovarian tumors than transabdo-minal scanning on the test set. For identifying the benign and malignant ovarian tumors containing cystic components, a mixture of transvaginal and transabdominal scanning increased the accuracy by 9.7% over transabdominal scanning alone. Conclusion CNN can identify ovarian malignant tumors under both scanning methods, but the accuracy of transvaginal scanning is higher than that of transabdominal scanning, and the CNN model has a higher accuracy in identifying benign and malignant ovarian tumors under transvaginal scanning.
