Ovarian Cancer Diagnosis and Risk Assessment

Ovarian cancer remains the most lethal gynaecological malignancy worldwide, largely owing to nonspecific symptoms and the absence of effective early‐stage screening. Accurate diagnosis and risk assessment hinge on the integration of clinical presentation, serum biomarkers and imaging modalities. Serum markers such as CA125 and HE4 assist in raising suspicion, yet both may be elevated in benign conditions or normal in early tumours. Ultrasonography remains the first‐line imaging technique, with expert assessment informing risk stratification. In recent years, structured reporting systems and multivariable risk models have been developed to classify adnexal masses, aiming to guide management and improve patient outcomes. Advances in artificial intelligence and multimodal algorithms have further enhanced diagnostic precision, offering potential to standardise interpretation, reduce inter‐observer variability and alleviate reliance on specialist expertise. Ongoing efforts focus on validating these tools across diverse populations and care settings, ensuring robust calibration and clinical utility before widespread adoption.

Research from Nature Portfolio

Recent studies have demonstrated transformative applications of deep learning to ultrasound imaging. A multicentre, international effort developed transformer‐based models trained on over 17 000 images from multiple countries, achieving superior sensitivity, specificity and overall accuracy compared with both expert and non-expert examiners. This AI-driven system generalises across different ultrasound platforms and patient demographics, and simulated triage scenarios suggest a substantial reduction in expert referrals while maintaining diagnostic performance. In parallel, an interpretable multimodal model named OvcaFinder combined deep‐learning predictions from ultrasound images with radiologist O-RADS scores and routine clinical data. OvcaFinder outperformed standalone models, achieving area under the curve values above 0.97 in internal datasets and 0.95 in external validation, and improved both the accuracy and consistency of radiologists’ assessments by reducing false positives and enhancing inter-reader agreement.

Ovarian Cancer Diagnosis and Risk Assessment publication trend

The graph below shows the total number of articles in ovarian cancer diagnosis and risk assessment across all publications each year (not limited to Nature Index journals).

Technical terms

CA125: A tumour-associated glycoprotein measured in serum that is often elevated in epithelial ovarian cancer but can rise in benign pelvic conditions.

HE4: A protein biomarker overexpressed in many ovarian tumours, offering improved specificity over CA125, particularly in the setting of endometriosis.

ADNEX model: A multivariable risk-prediction tool from the International Ovarian Tumour Analysis group that estimates the probability of benign versus various malignant adnexal masses.

O-RADS: A structured imaging lexicon that categorises adnexal masses by estimated risk of malignancy using standardised sonographic or MRI features.

Area under the curve (AUC): A summary measure of a diagnostic test’s discriminatory ability, reflecting the trade-off between sensitivity and specificity.

Sensitivity and specificity: Metrics indicating, respectively, the proportion of true positives correctly identified and the proportion of true negatives correctly excluded by a diagnostic test.

References

1. International multicenter validation of AI-driven ultrasound detection of ovarian cancer. Nature Medicine (2025).
2. Development and validation of an interpretable model integrating multimodal information for improving ovarian cancer diagnosis. Nature Communications (2024).
3. Risk-prediction models in postmenopausal patients with symptoms of suspected ovarian cancer in the UK (ROCkeTS): a multicentre, prospective diagnostic accuracy study. The Lancet Oncology (2024).
4. ADNEX risk prediction model for diagnosis of ovarian cancer: systematic review and meta-analysis of external validation studies. BMJ Medicine (2024).
5. External Validation of the Ovarian-Adnexal Reporting and Data System (O-RADS) Lexicon and the International Ovarian Tumor Analysis 2-Step Strategy to Stratify Ovarian Tumors Into O-RADS Risk Groups. JAMA Oncology (2023).
6. Biomarkers and algorithms for diagnosis of ovarian cancer: CA125, HE4, RMI and ROMA, a review. Journal of Ovarian Research (2019).