Background: Differentiating benign from malignant adnexal masses remains a common diagnostic challenge. While MRI is widely used when ultrasound is indeterminate, contrast-enhanced ultrasound (CEUS) provides real-time vascular assessment at the bedside. Comparative data between CEUS and MRI are limited in Indian clinical practice. Objective: To compare the diagnostic performance of CEUS and MRI for indeterminate adnexal masses, using histopathology and/or follow-up as reference standards. Methods: In this prospective study, 100 women with indeterminate adnexal masses underwent CEUS and MRI. Risk categories were assigned using a five-point CEUS score (malignant if ≥3) and O-RADS MRI (malignant if ≥4). Diagnostic accuracy, sensitivity, specificity, predictive values, and area under the ROC curve (AUC) were calculated. Results: The prevalence of malignancy was 27% (n=27). Both CEUS and MRI achieved 100% sensitivity and 100% NPV. MRI demonstrated higher specificity (84% vs 76%) and accuracy (88% vs 82%) compared with CEUS. AUCs confirmed excellent performance for both modalities (CEUS 0.981, MRI 0.965). Cohen’s κ indicated substantial agreement (0.65). False positives were more frequent with CEUS (n=18) than MRI (n=12), predominantly in borderline and complex benign cysts. Conclusion: CEUS and MRI both provide excellent discrimination of indeterminate adnexal masses. MRI achieved greater specificity and overall accuracy, while CEUS offered strong sensitivity and real-time utility. CEUS may serve as a rapid triage tool in resource-limited settings, with MRI retained as the definitive imaging standard.
Adnexal masses represent a common diagnostic challenge in gynaecologic imaging, with the critical clinical distinction lying between benign and malignant lesions. Although conventional ultrasound is the first-line modality, up to 20–30% of adnexal masses remain indeterminate on grayscale and Doppler assessment [1]. The accurate characterization of these indeterminate lesions is essential, as it directly influences patient counselling, surgical planning, and referral pathways.
Contrast-enhanced ultrasound (CEUS) has emerged as a valuable adjunct to conventional ultrasound. By providing real-time visualization of tumour vascularity and microcirculation, CEUS enhances the ability to distinguish malignant from benign adnexal lesions [2]. Meta-analyses have confirmed its diagnostic value, reporting pooled sensitivities exceeding 90% and specificities around 80% [1]. CEUS is also attractive in resource-constrained settings, given its wide availability, lack of ionizing radiation, and relatively low cost.
Magnetic resonance imaging (MRI), on the other hand, is widely accepted as the problem-solving modality for sonographically indeterminate adnexal masses. The European Society of Urogenital Radiology (ESUR) has issued comprehensive recommendations for its use, highlighting the value of multiparametric protocols including diffusion-weighted imaging and dynamic contrast enhancement [3]. The development of structured scoring systems such as the O-RADS MRI lexicon has further standardized interpretation and improved reproducibility.
Several multicentre studies have demonstrated that CEUS, when applied with structured criteria, significantly improves diagnostic confidence. Zhang et al. (2014) showed that qualitative CEUS assessment increased the sensitivity of adnexal mass characterization compared to conventional ultrasound alone [4]. More recently, Wang et al. (2023) compared O-RADS US combined with CEUS to the ADNEX MR scoring system, finding that the hybrid ultrasound approach provided diagnostic performance comparable to MRI [5]. Similarly, Xu et al. (2023) reported that integrating CEUS parameters into O-RADS US significantly enhanced specificity without compromising sensitivity [6].
Despite these advances, direct head-to-head comparisons of CEUS and MRI for indeterminate adnexal masses remain relatively scarce. Given differences in accessibility, cost, and expertise requirements, particularly in developing countries, such comparative studies are clinically important. The present prospective study therefore aimed to compare the diagnostic performance of CEUS and MRI in women with indeterminate adnexal masses at a tertiary referral centre in West Bengal, India, using histopathology and/or structured follow-up as the reference standard.
Objectives
The objective of this study was to compare the diagnostic performance of contrast-enhanced ultrasound (CEUS) and magnetic resonance imaging (MRI) in the characterization of indeterminate adnexal masses. Specifically, we aimed to evaluate and contrast their sensitivity, specificity, predictive values, overall accuracy, and area under the ROC curve, using histopathology and/or structured imaging follow-up as the reference standard.
Study design and setting
This was a prospective diagnostic accuracy study conducted in the Department of Radiodiagnosis, Burdwan Medical College, West Bengal, India, over a period of 12 months. Institutional Ethics Committee approval was obtained, and informed consent was secured from all participants prior to enrolment.
Study population
Consecutive women referred for evaluation of adnexal masses were screened.
Inclusion criteria:
Exclusion criteria:
A total of 100 women fulfilling the eligibility criteria were included.
Clinical and laboratory evaluation
Demographic details, menopausal status, serum CA-125 levels, and clinical presentation were recorded. Patients were categorized as pre- or postmenopausal according to standard definitions.
CEUS protocol
All examinations were performed using a high-end ultrasound system equipped with contrast-specific software and low mechanical index imaging. A bolus of 2.4 mL sulfur hexafluoride microbubbles was injected intravenously, followed by a saline flush. Continuous cine recording was acquired for up to 3 minutes to evaluate enhancement dynamics. Lesion characteristics assessed included:
A five-point CEUS scoring system was applied, with scores ≥3 considered malignant-positive.
MRI protocol
Pelvic MRI was performed on a 1.5T system using a standardized protocol including:
Interpretation was performed according to the O-RADS MRI scoring system (1–5), with scores ≥4 considered malignant-positive.
Reference standard
The reference standard was histopathology following surgical excision, available in 82 cases. In the remaining 18 cases, benign classification was confirmed by clinical and imaging follow-up for ≥6 months, documenting lesion stability or resolution.
Blinding
Two radiologists with >5 years of experience independently interpreted CEUS and MRI images. CEUS readers were blinded to MRI findings, and vice versa. Both were blinded to final outcomes. Disagreements were resolved by consensus.
Statistical analysis
Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated for CEUS and MRI. 95% confidence intervals (CIs) were derived using the Wilson method. Receiver operating characteristic (ROC) curves were generated, and the area under the curve (AUC) was calculated for each modality. Inter-modality agreement was assessed using Cohen’s κ statistic. Statistical significance was set at p < 0.05. Data were analysed using SPSS software (version 24.0; IBM Corp., Armonk, NY, USA) and R (R Foundation for Statistical Computing, Vienna, Austria).
A total of 100 women with indeterminate adnexal masses on baseline ultrasound were included. The prevalence of malignancy was 27% (n = 27). The median age of the cohort was 45 years (range: 18–78 years), with malignant cases being older than benign (median 50.0 vs 42.0 years). Overall, 34% of patients were postmenopausal, and postmenopausal status was more frequent among malignant cases (60% vs 30%).
The median lesion size was larger in malignant compared with benign masses (7.9 cm vs 6.2 cm, respectively). Serum CA-125 levels were also higher in malignant lesions (median 148.0 U/mL) compared to benign lesions (median 62.0 U/mL), though overlap was observed.
Morphological features suggestive of malignancy were enriched in the malignant group: papillary projections were present in 65% of malignant vs 25% of benign lesions, and ascites in 35% vs 15%, respectively.
Table 1 summarizes baseline characteristics stratified by final diagnosis.
Table 1. Baseline characteristics by final diagnosis
Characteristic |
Benign (n=73) |
Malignant (n=27) |
Median age (years) |
42.0 |
50.0 |
Postmenopausal (%) |
30% |
60% |
Median lesion size (cm) |
6.2 |
7.9 |
Median CA-125 (U/mL) |
62.0 |
148.0 |
Papillary projections (%) |
25% |
65% |
Ascites (%) |
15% |
35% |
Of the 100 adnexal masses, 73 (73%) were benign and 27 (27%) were malignant or borderline. Among the benign lesions, the most common subtypes were serous cystadenoma (24.7%), mature cystic teratoma (19.2%), and endometrioma (16.4%), followed by mucinous cystadenoma (13.7%) and a smaller proportion of other benign entities including fibroma/thecoma, haemorrhagic cyst, hydrosalpinx, and paraovarian cyst.
Within the malignant and borderline group, high-grade serous carcinoma was the predominant subtype (33.3%), with mucinous carcinoma (18.5%) and borderline serous or mucinous tumours (18.5%) also frequent. Less common histologies included endometrioid carcinoma (11.1%), clear cell carcinoma (7.4%), and metastatic carcinoma (11.1%).
The detailed histopathological distribution of adnexal mass subtypes is summarized in Table 2.
Table 2. Distribution of adnexal mass subtypes
Diagnosis category |
n (%) of total cohort |
Benign lesions (n=73) |
|
– Serous cystadenoma |
18 (24.7%) |
– Mature cystic teratoma |
14 (19.2%) |
– Endometrioma |
12 (16.4%) |
– Mucinous cystadenoma |
10 (13.7%) |
– Others* |
19 (26.0%) |
Malignant/borderline (n=27) |
|
– High-grade serous carcinoma |
9 (33.3%) |
– Mucinous carcinoma |
5 (18.5%) |
– Borderline tumours |
5 (18.5%) |
– Endometrioid carcinoma |
3 (11.1%) |
– Clear cell carcinoma |
2 (7.4%) |
– Metastatic carcinoma |
3 (11.1%) |
*Includes fibroma/thecoma, haemorrhagic cyst, hydrosalpinx, paraovarian cyst.
Both CEUS and MRI demonstrated excellent diagnostic ability in characterizing indeterminate adnexal masses. Using the predefined thresholds (CEUS score ≥3; O-RADS MRI ≥4), both modalities achieved 100% sensitivity, correctly identifying all malignant lesions (n = 27).
Specificity was higher for MRI compared with CEUS (84.0% vs 76.0%), which translated into greater overall accuracy (88.0% vs 82.0%). Positive predictive value was modest for both, reflecting the inclusion of complex benign lesions, but was higher with MRI (67.6% vs 58.1%). Negative predictive value was 100% for both modalities, underscoring their reliability in excluding malignancy.
Receiver operating characteristic (ROC) analysis confirmed excellent discrimination, with areas under the curve (AUC) of 0.981 for CEUS and 0.965 for MRI (Figure 1).
Table 3 summarizes the diagnostic performance metrics with 95% confidence intervals.
Table 3. Diagnostic performance of CEUS and MRI with 95% confidence intervals
Metric |
CEUS |
MRI |
Sensitivity |
100.0% (86.7–100.0) |
100.0% (86.7–100.0) |
Specificity |
76.0% (65.2–84.2) |
84.0% (74.1–90.6) |
PPV |
58.1% (43.3–71.6) |
67.6% (51.5–80.4) |
NPV |
100.0% (93.7–100.0) |
100.0% (94.3–100.0) |
Accuracy |
82.0% (73.3–88.3) |
88.0% (80.2–93.0) |
Figure 2. Comparative diagnostic performance of CEUS and MRI. Bar chart illustrating specificity and overall accuracy for both modalities. MRI demonstrated higher specificity (84% vs 76%) and accuracy (88% vs 82%) compared with CEUS, while both modalities achieved identical sensitivity (100%) and negative predictive value (100%).
The confusion matrices for CEUS and MRI are presented in Table 4. Both modalities correctly classified all malignant lesions, yielding no false negatives. The primary source of error arose from false-positive classifications of benign lesions.
For CEUS, 18 benign lesions were incorrectly classified as malignant, compared with 12 for MRI. Many of these false positives corresponded to benign complex cysts and borderline tumours, which often demonstrated suspicious vascularity on CEUS or overlapping O-RADS features on MRI. True negatives were therefore more frequent with MRI (n = 63) than with CEUS (n = 57).
Overall, these patterns explain the higher specificity and accuracy observed for MRI, despite both modalities achieving identical sensitivity.
Table 4. Confusion matrices for CEUS (threshold ≥3) and MRI (O-RADS ≥4)
|
CEUS: Actual Benign |
CEUS: Actual Malignant |
Predicted Negative |
57 |
0 |
Predicted Positive |
18 |
25 |
|
CEUS: Actual Benign |
CEUS: Actual Malignant |
Predicted Negative |
57 |
0 |
Note: Counts are for the full cohort (N=100). Thresholds were prespecified.
Receiver operating characteristic (ROC) analysis demonstrated excellent discrimination for both CEUS and MRI (Figure 1). The area under the curve (AUC) was 0.981 for CEUS and 0.965 for MRI, indicating high diagnostic accuracy for both modalities. The ROC curves were closely aligned, reflecting comparable overall performance, although MRI offered slightly higher specificity at the predefined diagnostic threshold.
Figure 1. ROC curves comparing CEUS and MRI in the diagnosis of indeterminate adnexal masses.
Agreement between CEUS and MRI in dichotomized classification (positive vs negative) was substantial, with a Cohen’s κ of 0.65. Discordant cases were predominantly benign lesions that CEUS classified as malignant but MRI classified correctly, such as complex cystadenomas and endometriomas with vascular enhancement on CEUS but non-suspicious features on MRI. A smaller number of cases demonstrated the reverse pattern, where MRI yielded false-positive O-RADS scores in borderline or inflammatory lesions.
Key Takeaway
In summary, both CEUS and MRI demonstrated excellent diagnostic performance in the evaluation of indeterminate adnexal masses, achieving perfect sensitivity and very high negative predictive value. MRI provided superior specificity and overall accuracy, while CEUS maintained strong performance as a rapid, bedside tool. Together, these findings support a complementary role for CEUS as a triage modality and MRI as the definitive imaging standard when available.
In this prospective comparative study of 100 women with indeterminate adnexal masses, both CEUS and MRI demonstrated excellent diagnostic performance, each achieving 100% sensitivity and NPV, while MRI achieved higher specificity (84% vs 76%) and overall accuracy (88% vs 82%) compared with CEUS. These findings align with and expand upon a growing body of literature exploring the complementary roles of CEUS and MRI in adnexal mass characterization.
Shi et al. (2023) reported that combining O-RADS US with CEUS significantly improved risk stratification of adnexal masses compared to O-RADS alone, with sensitivity approaching 95% and specificity around 82% [7]. Our study reinforces this observation by demonstrating that CEUS alone, when applied with structured scoring, achieves comparable sensitivity and acceptable specificity. Jiang et al. (2025) further advanced this integration by incorporating O-RADS US v2022, CEUS, and CA-125 into the OCC-US model, achieving a sensitivity of 98% and a specificity of 87% [8]. These values are slightly higher than those observed in our CEUS analysis (100% sensitivity, 76% specificity) and comparable to our MRI performance (100% sensitivity, 84% specificity). This supports the concept that biomarker integration can meaningfully reduce the false-positive rates we observed with CEUS (n = 18).
Similarly, Zhang et al. (2025) confirmed that the combination of O-RADS, CEUS, and CA-125 improved malignancy identification, particularly for borderline tumours [9]. In our series, borderline lesions accounted for a portion of CEUS false positives, suggesting that multimodal diagnostic algorithms could mitigate this challenge. Shentu et al. (2022) demonstrated that CEUS effectively distinguishes benign from malignant adnexal lesions, with an AUC of 0.92 [10], consistent with the AUC of 0.981 achieved in our CEUS cohort.
MRI performance in our study closely parallels prior reports. Sadowski et al. (2022), in their description of the O-RADS MRI system, emphasized its high sensitivity (93–98%) and strong reproducibility [11]. Our MRI results (sensitivity 100%, specificity 84%) are consistent with these benchmarks, although our specificity was slightly higher than that reported in multicentre evaluations. Likewise, Guo et al. (2021) showed that structured MRI-based systems such as GI-RADS could achieve specificities of 80–85% [12], again comparable to our findings.
CEUS continues to show added value in specific contexts. Lu et al. (2024) demonstrated that CEUS, when combined with IOTA simple rules and O-RADS, yielded superior performance for masses with solid components [13]. This resonates with our dataset, where papillary projections were present in 65% of malignant lesions and frequently triggered CEUS suspicion, though sometimes in benign cystadenomas. Wu et al. (2024) compared visual CEUS assessment with time–intensity curve analysis, showing improved reproducibility with quantitative approaches [14]. Our reliance on categorical scoring reflects current practice, but the incorporation of TICs may further reduce subjectivity.
Meta-analyses reinforce these trends. Rizzo et al. (2022) systematically reviewed O-RADS MRI studies across 11 cohorts and reported pooled sensitivity of 93% and specificity of 81%, with malignancy rates in O-RADS 5 lesions exceeding 95% [15]. Our MRI results (sensitivity 100%, specificity 84%) fall well within this performance range, suggesting that the slightly higher sensitivity in our cohort may reflect sample size and referral bias, while specificity remained consistent with pooled estimates.
Complementary innovations are ongoing in CEUS. Delaney et al. (2020) piloted contrast-enhanced subharmonic imaging in 42 adnexal lesions and reported an AUC of 0.90, with sensitivity 92% and specificity 78% [16]. These values are highly comparable to our CEUS AUC of 0.981, sensitivity 100%, and specificity 76%, underscoring both the promise and the consistency of vascular-based CEUS techniques across platforms.
Finally, broader contextual studies remind us of the clinical stakes. Sadowski et al. (2018) reported that up to 15% of adnexal cysts classified as indeterminate at baseline ultrasound ultimately represented malignancy [17]. Our observed malignancy prevalence of 27% is higher, reflecting the referral bias of a tertiary Indian centre. This highlights how regional and institutional variations, including patient selection, imaging equipment, and reader experience, influence diagnostic yields.
Taken together, our findings situate CEUS and MRI within a complementary diagnostic paradigm. CEUS provides real-time, bedside vascular characterization with strong sensitivity, while MRI offers higher specificity and accuracy, particularly when structured systems like O-RADS are applied. The integration of these modalities with biomarker assays, as suggested in recent literature [7–9], may represent the most effective pathway to minimize both false negatives and false positives across diverse practice environments.
This single-centre study had a modest sample size, and the 27% malignancy prevalence reflects referral bias, which may limit generalizability. Some benign cases were classified by follow-up rather than histology, introducing potential verification bias. CEUS was assessed by categorical scoring without time–intensity curve analysis, and both modalities are operator- and equipment-dependent. Finally, integrated multimodal approaches such as OCC-US [8] were not evaluated, though they may further improve specificity.
In women with indeterminate adnexal masses, both CEUS and MRI demonstrated excellent diagnostic performance, achieving perfect sensitivity and very high NPV. MRI offered superior specificity and overall accuracy, while CEUS provided strong real-time assessment at the point of care. These findings support the use of CEUS as a cost-effective triage tool in resource-limited settings, with MRI reserved for problem-solving and staging. Integrating imaging with structured risk systems and biomarkers, as highlighted in recent literature [7–9], may further refine risk stratification and reduce false positives.