Thyroid nodules are frequently detected, particularly through imaging, yet only a small proportion are malignant or symptomatic. Guidelines published in the European Journal of Ultrasound address the persistent challenges this creates for diagnosis and management, as many patients still undergo repeated follow-up examinations and invasive procedures, including fine-needle aspiration and surgery. Advances in ultrasound technology, including contrast-enhanced imaging, elastography and artificial intelligence, support a broader assessment approach. Multiparametric ultrasound brings these tools together to improve characterisation, reduce unnecessary interventions and support clinical decision-making across a range of thyroid conditions.
Risk Stratification Systems and Diagnostic Performance
Ultrasound remains the primary modality for initial assessment and long-term monitoring of thyroid nodules due to its accessibility and high resolution. Traditional reliance on individual ultrasound features, such as echogenicity or calcifications, shows limitations related to observer variability. To address this, risk stratification systems based on combined imaging features have been developed, commonly grouped under Thyroid Imaging Reporting and Data Systems. These systems provide structured frameworks to estimate malignancy risk and support decisions on further investigation.
Different TIRADS models vary in structure, ranging from pattern-based classifications to point-based scoring systems. Their implementation aims to standardise reporting and improve diagnostic consistency. Comparative evaluations show broadly similar performance across major systems, although variations exist in sensitivity and specificity. Some systems demonstrate stronger sensitivity, while others achieve higher specificity, reflecting trade-offs in clinical application. Combining multiple systems may improve diagnostic accuracy in selected cases, particularly when results are discordant.
Must Read: Deep Learning Advances in Ultrasound Segmentation
Despite encouraging results, challenges remain in routine implementation. Variability in interpretation and workflow integration can limit adoption. Structured reporting templates, targeted education and decision-support tools at the point of care support wider use. Certain nodule types, such as autonomously functioning nodules, present limitations for these systems due to overlapping imaging features with malignant lesions. Clinical context therefore remains essential to avoid unnecessary procedures. Consistent application of a single stratification system during follow-up is recommended to maintain continuity in patient management.
Multiparametric Ultrasound in Indeterminate Nodules
Cytologically indeterminate thyroid nodules represent a significant diagnostic challenge, with a notable proportion associated with malignancy. Conventional ultrasound alone shows limited reliability in distinguishing benign from malignant lesions in this group, partly due to overlapping imaging characteristics and inherent biases in available evidence. The heterogeneity of indeterminate nodules, including entities with differing biological behaviour, further complicates risk assessment.
Multiparametric ultrasound expands diagnostic capability by incorporating additional imaging modalities. Elastography, which assesses tissue stiffness, contributes valuable information in this setting. Soft nodules show a high negative predictive value for malignancy, supporting more conservative management in selected cases. Quantitative and semi-quantitative elastographic techniques demonstrate improved diagnostic accuracy compared with qualitative approaches, although performance varies depending on methodology.
Systematic analyses indicate moderate sensitivity and specificity for elastography in indeterminate nodules, with shear wave techniques providing consistent results. However, limitations persist due to retrospective study designs and selection bias, particularly in cohorts restricted to surgical cases. These constraints affect the ability to generalise findings and limit the identification of specific tumour subtypes. Nevertheless, combining elastography with structured risk stratification systems supports more refined decision-making and may reduce unnecessary repeat biopsies or surgical interventions.
Artificial Intelligence and Emerging Technologies
Artificial intelligence introduces additional capabilities for thyroid nodule evaluation through radiomics and computer-assisted diagnostic systems. These approaches analyse imaging data to extract quantitative features and support classification. Evidence indicates that AI-based tools achieve diagnostic performance comparable to experienced operators in characterising thyroid nodules. This level of performance highlights their potential role in standardising assessments and supporting less experienced clinicians.
Integration of AI with established risk stratification systems aims to enhance consistency and improve overall diagnostic accuracy. Variability in human interpretation remains a recognised limitation of ultrasound-based assessment, and automated analysis offers a pathway to reduce this variability. AI-assisted systems contribute to more consistent evaluation of imaging features, potentially improving risk stratification and management decisions.
Despite promising results, current evidence reflects substantial heterogeneity in algorithm design and application. Studies often focus on specific implementations rather than generalisable frameworks, limiting direct comparison across systems. Regulatory-approved tools demonstrate similar sensitivity to radiologists, but broader validation across clinical settings remains limited. As a result, AI remains primarily a research tool in this domain.
Ongoing development focuses on improving integration with clinical workflows and expanding training datasets to enhance robustness. While AI-assisted evaluation shows potential benefits, routine use in clinical practice is not yet supported. Further evidence is required to confirm its reliability, reproducibility and impact on patient outcomes before widespread adoption.
Multiparametric ultrasound provides a structured approach to thyroid nodule evaluation, integrating multiple imaging techniques to improve diagnostic precision. Risk stratification systems standardise assessment but require consistent application and clinical context for optimal use. Elastography enhances characterisation of indeterminate nodules and supports more selective intervention strategies. Artificial intelligence introduces additional analytical capacity, with performance comparable to experienced clinicians, although current use remains limited to research settings. Combined application of these tools supports more targeted management, aiming to reduce unnecessary procedures while maintaining diagnostic accuracy.
Source: European Journal of Ultrasound
Image Credit: iStock
References:
Cantisani V, Radzina M, Dietrich CF (2026) EFSUMB Guidelines on Multiparametric Ultrasound Thyroid Nodule Evaluation: PART I. Ultraschall Med: eFirst.
Cantisani V, Radzina M, Dietrich CF (2026) EFSUMB Guidelines on Multiparametric Ultrasound Thyroid Nodule Evaluation: PART II. Ultraschall Med: eFirst.
