While thyroid nodules are detected in up to 68% of clinical practice, malignant cases consistently remain between 5% and 15%. A Bayesian network meta-analysis published in the journal Frontiers in Endocrinology reveals that combining contrast-enhanced ultrasound (CEUS) with elastography significantly improves diagnostic accuracy, achieving an area under the curve (AUC) of 0.97 and outperforming contrast imaging used in isolation.
Combining CEUS with Elastography Improves Thyroid Nodule Diagnosis
- What was tested: Researchers analyzed whether pairing contrast-enhanced ultrasound (CEUS) with other diagnostic tools—like elastography or artificial intelligence—does a better job of telling benign thyroid nodules apart from cancerous ones.
- The core finding: Combining CEUS with elastography yielded the highest overall diagnostic performance, spotting malignant nodules with 92% sensitivity.
Bayesian Network Meta-Analysis Evaluates Diagnostic Precision
The systematic review searched databases including PubMed, Web of Science, Embase, and Cochrane up to October 15, 2025. Out of the evaluated literature, 24 studies originated in China, 19 operated as retrospective cohorts, and 17 achieved high-quality ratings on the QUADAS-2 assessment tool.
Standalone CEUS relies on intravenously injected microbubbles to display dynamic microvascular perfusion patterns qualitatively and quantitatively. However, overlapping enhancement patterns between benign and malignant lesions limit its standalone utility, typically restricting its role to a supplement to conventional ultrasound. To resolve inconsistencies among prior combined strategies, the research team deployed a Bayesian network meta-analysis following PRISMA guidelines.
Performance Metrics of Combined Diagnostic Strategies
When evaluated independently, CEUS yielded an aggregated sensitivity of 0.83, a specificity of 0.81, and an AUC of 0.91. Every paired combination tested in the review demonstrated superior diagnostic accuracy over the contrast method used alone. The pairing of CEUS with elastography—a technique measuring tissue stiffness—secured the highest superiority index at 3.47, a sensitivity of 0.92, and an AUC scaling from 0.96 to 0.99.
Other combinations also exhibited distinct advantages. Integrating CEUS with artificial intelligence (AI) reached a sensitivity of 0.93. Meanwhile, pairing CEUS with conventional ultrasound registered the highest specificity at 0.90 and a diagnostic odds ratio (DOR) of 78.89, compared to just 21.73 for CEUS alone. Heterogeneity across the studies remained moderate, registering an I² of 61% for sensitivity and 60.3% for specificity.
| Diagnostic Strategy | Pooled Sensitivity (95% CI) | Pooled Specificity (95% CI) | Area Under Curve (AUC) |
|---|---|---|---|
| CEUS Alone | 0.83 (0.79–0.86) | 0.81 (0.77–0.85) | 0.91 (0.89–0.93) |
| CEUS + Elastography | 0.92 (0.88–0.95) | 0.83 (0.75–0.89) | 0.97 (0.96–0.99) |
| CEUS + TI-RADS | 0.92 (0.89–0.94) | 0.80 | Not Reported |
| CEUS + Conventional Ultrasound | 0.89 (0.74–0.97) | 0.90 (0.83–0.95) | Not Reported |
AI Integration Boosts Sensitivity in High Risk Cohorts
Subgroup evaluations highlighted the utility of machine learning integration within elevated-risk cohorts. Across 18 studies examining high-risk populations—characterized by TI-RADS 4–5 classifications, Bethesda V–VI cytology, suspicious signs, or scheduled surgeries—the association between CEUS and AI achieved a peak sensitivity of 0.95 and a SUCRA ranking of 90%. Conversely, in eight studies lacking risk stratification, the integration of conventional ultrasound led performance.
Important methodological limitations acknowledged by the study’s authors include variations in CEUS protocols, elastography thresholds, and AI algorithms. The strong concentration of data originating from Asian cohorts necessitates independent validation across non-Asian populations before global clinical guidelines can adapt these combined strategies.
Individualizing Thyroid Imaging Protocols
The findings emphasize that while CEUS retains standalone merit, merging structural, mechanical, and automated morphological data substantially sharpens clinical decision-making. Tailoring diagnostic protocols to individual patient profiles relies heavily on local equipment availability and operator expertise. As multi-modal imaging continues to evolve, large-scale prospective trials will ultimately define standardized pathways for thyroid cancer detection.
