Ricerc@Sapienza

Thyroid nodules are a very common finding, with a prevalence greater than 50% in women and older populations (>60 years old). The clinical impact is primarily represented by the precocious identification of thyroid nodule malignant potential, representing the 5-10% of thyroid nodule. Currently the gold standard for the diagnosis of thyroid tumors is fine needle aspiration cytology (FNAC), but in the last decades there have been many efforts in order to reduce the number of unnecessary procedures by enhancing the diagnostic accuracy of ultrasound. Ultrasonography (US) is the gold standard imaging in thyroid diseases, offering many perks such as being economic and non invasive, besides having a wide diffusion on the territory. The improvements in this field have been aimed on: 1) developing new techniques in order to better characterize thyroid nodules, such as elastography; 2) identifying the neoplastic iconographic features to create classification systems able to stratify nodules on a risk basis, such as TI-RADS; 3) create automated
systems based on artificial intelligence to help physician in detecting and evaluating thyroid nodules, on the model of Computer Aided Diagnosis (CAD) systems already existing in other radiological fields.
The aim of the present project is therefore to assess the diagnostic accuracy and interobserver variability of ultrasound techniques performed by physicians with different levels of expertise using two distinct TI-RADS and elastography in comparison with two CAD systems, and to create a repository for newer AI System.