Blood Biomarkers Boost Ultrasound’s Power to Spot Ovarian Cancer Early

Ovarian cancer has long been one of medicine’s most elusive targets. It is the fifth leading cause of cancer-related death among women in the Western world, yet more than 70 percent of cases are still diagnosed at advanced Stage III or IV disease, when the tumor has already spread beyond the ovary. Mortality rates have barely budged in three decades, not because treatments have stood still—aggressive cytoreductive surgery and novel chemotherapies have advanced considerably—but because the disease whispers before it screams. Its earliest symptoms are subtle, and reliable early detection methods remain frustratingly absent. Now, a team of Chinese researchers reports that a simple modification to the most widely used ultrasound classification system could sharpen the line between benign and malignant ovarian masses, potentially sparing hundreds of thousands of women worldwide from unnecessary surgery while fast-tracking those who truly need urgent care.

The study, published in Frontiers in Oncology by Mengze Liu, Dandan Li of Weifang People’s Hospital and Zhanqiang Jin of the University of South China, tackles a well-known weakness in the American College of Radiology’s Ovarian-Adnexal Reporting and Data System for ultrasound, known as O-RADS US. That framework asks sonographers to describe an ovarian mass in standardized terms—whether it is a simple cyst or a complex lesion, whether its walls are smooth or irregular, whether it contains solid components or papillary projections, and how much blood flow it shows on Doppler imaging—and then assigns the lesion to a category that carries an estimated risk of malignancy. Category 2 lesions carry less than a 1 percent risk, Category 3 lesions between 1 and 10 percent, Category 4 lesions between 10 and 50 percent, and Category 5 lesions 50 percent or more. The problem is that Category 4, the intermediate zone, spans an enormous range of risk, and the entire system depends on the subjective judgment and experience of the person holding the ultrasound probe.

To inject objectivity into that gray zone, the researchers turned to blood. They integrated the Risk of Ovarian Malignancy Algorithm, or ROMA, a U.S. Food and Drug Administration-approved calculation that combines two serum biomarkers—cancer antigen 125 (CA125) and human epididymis protein 4 (HE4)—with the patient’s menopausal status. CA125 has been a staple of ovarian cancer diagnostics for decades, but it rises in pregnancy, menstruation, endometriosis and inflammatory pelvic disease, giving it poor specificity. HE4, a more recently adopted marker, complements CA125 because it is less often elevated in those benign conditions, though smoking and contraceptive use can shift its levels too. Neither marker alone is sensitive or specific enough for early detection, but evidence has consistently supported pairing them. ROMA packages that pairing into a single probability score, adjusted for the fact that postmenopausal women face inherently higher baseline risk.

The team’s hybrid approach was elegantly simple. They first classified 159 ovarian-adnexal masses in 151 patients using standard O-RADS US criteria, with two sonographers of more than ten years’ experience each independently reviewing the images, blinded to all biomarker results, and a third senior examiner resolving disagreements. They then measured CA125 and HE4 in preoperative blood samples, calculated each patient’s ROMA value, and compared it against a diagnostic threshold derived from receiver operating characteristic analysis: 29.70 for premenopausal women and 42.50 for postmenopausal women. Depending on how far a patient’s ROMA score exceeded that threshold, the lesion’s O-RADS category was nudged up or down. A ROMA value exceeding the threshold by 10 percent or more escalated a Category 3 or 4 lesion to Category 4 or 5, while values below the threshold could demote intermediate-risk lesions into safer categories.

The results were striking. The modified system achieved an area under the receiver operating characteristic curve—a standard measure of diagnostic discrimination where 1.0 is perfect—of 0.922, significantly better than the original O-RADS US score of 0.815 and better than ROMA alone in any patient group. In practical terms, the malignancy rate among lesions classified as high-risk (Categories 4 and 5) rose from 44.0 percent under the original system to 58.1 percent under the modified one, meaning the high-risk pool became far more concentrated with true cancers. Meanwhile, the proportion of lesions languishing in the ambiguous Category 4 shrank from 29.5 percent to 15.1 percent, as many were confidently reassigned to lower-risk Categories 2 or 3. Category 5 lesions, the most worrisome, remained essentially stable at around 23 to 24 percent, suggesting the biomarker integration did not destabilize the system’s most confident calls.

The reclassification patterns reveal how the hybrid system behaves at the level of individual patients. Of the 47 lesions originally placed in Category 4, 26 were downgraded to Category 3 and only one of those proved malignant on final histopathology, while 14 were upgraded to Category 5 and seven of those were indeed cancers. Among the 30 original Category 3 lesions, all 20 that were downgraded to Category 2 turned out to be completely benign. The trade-off came in sensitivity: the modified system detected 82.1 percent of cancers compared with 92.3 percent for the original, but its specificity jumped from 60.0 percent to 92.5 percent. In a clinical context where false positives trigger surgery and false negatives can be catastrophic, that rebalancing matters enormously—it means far fewer benign masses would be sent down the invasive diagnostic pathway.

The study cohort itself reflected the diagnostic challenge. Of the 159 masses, 111 were benign—ranging from serous and mucinous cystadenomas to endometriosis cysts, fibromas and benign teratomas—11 were borderline tumors, and 37 were malignant, including high-grade and low-grade serous carcinomas, mucinous carcinomas, clear cell carcinoma and endometrioid carcinoma. Malignant lesions were also significantly larger, with a median diameter of 11.0 centimeters compared with 7.35 centimeters for benign masses, and carried markedly higher levels of CA125, HE4 and ROMA scores. The ROMA thresholds identified in this cohort yielded sensitivities and specificities of 80.6 percent and 76.8 percent in premenopausal women and 88.0 percent and 78.5 percent in postmenopausal women, broadly consistent with the 76 to 86 percent sensitivity and 74 to 95 percent specificity reported in prior meta-analyses.

The significance of this work lies in its attempt to fuse two diagnostic philosophies that have historically operated in parallel. O-RADS US captures the morphology and vascularity of a lesion—what the tumor looks like—while ROMA captures its molecular signature—what the tumor is secreting. A recent meta-analysis found the original O-RADS US system achieves a robust pooled sensitivity of 97 percent but only moderate specificity of 77 percent, precisely the profile that generates anxiety-inducing intermediate classifications and unnecessary operations. By letting a validated biomarker algorithm arbitrate the ambiguous middle of the risk spectrum, the modified framework converts subjective uncertainty into quantified probability. The authors suggest this could reduce unnecessary biopsies and surgeries for low-to-moderate risk lesions, a meaningful benefit given the physical, emotional and financial costs of adnexal surgery.

Important caveats temper the enthusiasm. The study was retrospective and single-center, and the ROMA thresholds were derived from the same patient cohort in which the new system was then tested—a methodological circularity that risks overfitting and an overly optimistic estimate of accuracy. Classic benign lesions such as simple cysts were excluded, introducing selection bias, and there was no independent validation cohort or formal interobserver assessment. The authors themselves call for prospective observational studies across diverse populations before the modified system enters clinical practice. Still, the concept is compelling: if larger, multi-center trials confirm these findings, the humble combination of an ultrasound examination and two blood tests could become a markedly more reliable gatekeeper, catching more of the ovarian cancers that matter while letting benign masses rest undisturbed.

Subject of Research: Integration of the ROMA biomarker algorithm into the O-RADS ultrasound system for risk stratification of ovarian carcinoma

Article Title: A modified ovarian-adnexal reporting and data system ultrasound for differentiating ovarian carcinoma: integration of the risk of malignancy algorithm into the American College of Radiology’s O-RADS US framework

Article References: Liu, M., Li, D., & Jin, Z. (2026). A modified ovarian-adnexal reporting and data system ultrasound for differentiating ovarian carcinoma: integration of the risk of malignancy algorithm into the American College of Radiology’s O-RADS US framework. Frontiers in Oncology, 16, Article 1962862. https://doi.org/10.3389/fonc.2026.1962862

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1962862

Keywords: ovarian cancer, O-RADS US, ROMA, CA125, HE4, transvaginal ultrasound, biomarkers, risk stratification, diagnostic imaging, gynecologic oncology, adnexal masses, early detection