Diagnostic Accuracy of Artificial Intelligence Models for Ovarian Tumor Classification on Ultrasound: A Systematic Review
Article Sidebar
Main Article Content
Abstract
Background: Artificial intelligence (AI) may support ultrasound-based characterization of ovarian and adnexal masses, but differences in diagnostic targets, validation methods, and reporting complicate interpretation of performance. Objective: To systematically evaluate the diagnostic accuracy of ultrasound-based AI models for distinguishing benign from malignant ovarian and adnexal masses and examine factors affecting clinical applicability. Methods: PubMed/MEDLINE, Embase, Scopus, and Web of Science were searched for publications from January 2022 through January 2026, with a search update in March 2026. Eligibility specified adult women, ultrasound-based AI classification, clinical comparators, and histopathological verification. Two reviewers independently screened records and assessed full texts. QUADAS-2 appraisal was described, and findings were synthesized narratively without statistical pooling. Results: The search identified 720 records, with 425 remaining after deduplication. Six reports were listed, covering deep learning pipelines, convolutional neural networks, hybrid echogenic component analysis, and multiclass classification. However, one February 2026 publication exceeded the stated eligibility window, and the eligibility of a cyst-subtype classification study remained unresolved. Participant totals, study-level diagnostic estimates, confidence intervals, and QUADAS-2 judgments were not presented. Consequently, diagnostic performance magnitude, precision, comparative superiority, and overall evidence reliability could not be established. Reported workflow and safety benefits lacked extracted outcome data. Conclusion: The assembled literature illustrates several AI approaches to ovarian ultrasound classification, but the present synthesis does not establish their diagnostic accuracy or readiness for routine implementation. Eligibility reconciliation, complete quantitative extraction, and transparent risk-of-bias reporting are necessary before clinical recommendations can be supported. Prospective independent validation should evaluate defined diagnostic tasks and patient-level consequences
Article Details
Issue
Section
This work is licensed under a Creative Commons Attribution 4.0 International License.
How to Cite
References
1. Dai WL, Wu YN, Ling YT, Zhao J, Zhang S, Gu ZW, et al. Development and validation of a deep learning pipeline to diagnose ovarian masses using ultrasound screening: a retrospective multicenter study. EClinicalMedicine. 2024;78.
2. Giourga M, Petropoulos I, Stavros S, Potiris A, Gerede A, Sapantzoglou I, et al. Enhancing ovarian tumor diagnosis: performance of convolutional neural networks in classifying ovarian masses using ultrasound images. J Clin Med. 2024;13(14):4123.
3. Li J, Chen Y, Zhang M, Zhang P, He K, Yan F, et al. A deep learning model system for diagnosis and management of adnexal masses. Cancers (Basel). 2022;14(21):5291.
4. Wu Y, Dai W, Li X, Zhang S, Gong L, Wang J, et al. Development and validation of an artificial intelligence-based model for diagnosing benign, borderline, and malignant adnexal masses. NPJ Precis Oncol. 2026;10(1):106.
5. Yoeli-Bik R, Whitney HM, Li H, Bilecz A, Abramowicz JS, Lan L, et al. Hybrid artificial intelligence echogenic components-based diagnosis of adnexal masses on ultrasound. Med Phys. 2025;52(7):e17983.
6. Parekh A, Desai M. EfficientOvaNet: efficient deep learning model for multiclass classification of benign ovarian cyst using ultrasound images. Neural Comput Appl. 2025;37(30):25495-514.