Abstract
Thyroid nodules are common incidental findings. Most of them are benign, but many
unnecessary fine-needle aspiration procedures, core biopsies, and even thyroidectomies
or non-invasive treatments have been performed. To improve thyroid nodule characterization,
the use of multiparametric ultrasound evaluation has been encouraged by most experts
and several societies. In particular, US elastography for assessing tissue stiffness
and CEUS for providing insight into vascularization contribute to improved characterization.
Moreover, the application of AI, particularly machine learning and deep learning,
enhances diagnostic accuracy. Furthermore, AI-based computer-aided diagnosis (CAD)
systems, integrated into the diagnostic process, aid in risk stratification and minimize
unnecessary interventions. Despite these advancements, challenges persist, including
the need for standardized TIRADS, the role of US elastography in routine practice,
and the integration of AI into clinical protocols. However, the integration of clinical
information, laboratory information, and multiparametric ultrasound features remains
crucial for minimizing unnecessary interventions and guiding appropriate treatments.
In conclusion, ultrasound plays a pivotal role in thyroid nodule management. Open
questions regarding TIRADS selection, consistent use of US elastography, and the role
of AI-based techniques underscore the need for ongoing research. Nonetheless, a comprehensive
approach combining clinical, laboratory, and ultrasound data is recommended to minimize
unnecessary interventions and treatments.
Keywords Thyroid nodule characterization - US-Elastography - TIRADS - CEUS - AI
