Search

Machine learning (ML) is an interdisciplinary sector in the subset of artificial intelligence (AI) that creates systems to set up logical connections using algorithms, and thus offers predictions for complex data analysis. In the present review, an up-to-date summary of the current state of the art regarding ML and AI implementation for thyroid nodule ultrasound characterization and cancer is provided, highlighting controversies over AI application as well as possible benefits of ML, such as, for example, training purposes. There is evidence that AI increases diagnostic accuracy and significantly limits inter-observer variability by using standardized mathematical algorithms. It could also be of aid in practice settings with limited sub-specialty expertise, offering a second opinion by means of radiomics and computer-assisted diagnosis. The introduction of AI represents a revolutionary event in thyroid nodule evaluation, but key issues for further implementation include integration with radiologist expertise, impact on workflow and efficiency, and performance monitoring.

Family history of thyroid cancer increases the risk of harboring thyroid malignancies that end up having extrathyroidal extension (ETE) and multifocality on histology, some authors suggest a more aggressive surgical approach. Their pre-operative identification could allow more conservative surgical procedures if none of these features are suspected. Our aim was to assess if neck ultrasonography could identify or exclude multifocality or ETE in these patients to tailor the extent of surgery. This retrospective study included patients with previous thyroid surgery, ≥1 first-grade relative with thyroid cancer, and who had undergone pre-surgical ultrasound. ETE was suspected in the case of thyroid border interruption or gross invasion of perithyroidal tissues. Multiple suspicious nodules were defined as suspicion of multifocal cancer. The cohort consisted of 45 patients (median age 49 years, 40 with thyroid cancer, 30 females). The positive predictive value of ultrasonography in predicting multifocality and ETE was 57.14% (25.25–84.03) and 41.67% (21.5–65.1%), respectively, while the negative predictive values were 63.2% (56.4–69.4%) and 72.7% (63.3–80.5%). Pre-operative ultrasound examination is unable to reliably identify or exclude multifocal disease or extrathyroidal extension. In patients scheduled for surgery and with a first-degree relative affected by DTC, a “negative” pre-operative US report does not exclude the potential finding of multifocality and ETE at final histopathology.

Ultrasound (US) is the first imaging modality for thyroid parenchyma evaluation. In the last decades, the role of ultrasound has been improved with the introduction of new US software, such as contrast-enhanced ultrasound (CEUS) and US-elastography (USE). USE is nowadays recognized as an essential part of the multiparametric ultrasound (MPUS) examination, in particular for the indeterminate thyroid nodule with possible fine-needle aspiration cytology (FNAC) number reduction; even if further and larger studies are needed to validate it. More controversial is the role of CEUS in thyroid evaluation, due to its high variability in sensitivity and specificity. Semi-automatic US systems based on the computer-aided diagnosis (CAD) system are producing interesting results, especially as an aid to less experienced operators. New knowledge on the molecular mechanisms involved in thyroid cancer is allowing practitioners to identify new genomic thyroid markers that could reduce the number of “diagnostic” thyroidectomies. We have therefore drawn up an updated representation of the current evidence in the literature for thyroid nodule multiparametric ultrasound (MPUS) evaluation with particular regard to USE, the US CAD system and CEUS.

Ultrasonographic risk-stratification systems (RSS), including various Thyroid Imaging Reporting and Data Systems (TIRADS), were proposed to improve reporting and reduce the number of fine-needle aspiration biopsies. However, age might be a confounder since some suspicious ultrasonographic features lack specificity in elderly patients. We aimed to investigate whether the diagnostic performance of the RSS varied between age groups. All patients consecutively referred for thyroid biopsy between November 1, 2015, and March 10, 2020, were included. The malignancy risk of each nodule was estimated according to five RSS: the American Association of Clinical Endocrinologists/American College of Endocrinology/Associazione Medici Endocrinologi guidelines, the American College of Radiology (ACR) TIRADS, the American Thyroid Association guidelines, the European TIRADS, and the Korean TIRADS. Overall, 818 nodules (57 malignant) were evaluated. The malignancy rate was higher in patients &le, 65 years (8.1%) than in patients &gt, 65 years (3.8%, p = 0.02). All RSS confirmed a significant discriminative performance in both age groups, with a negative predictive value of 100% in patients &gt, 65 years, although specificity was lower in older patients. The ACR TIRADS was the best performing in both age groups. RSS could avoid a sizable number of biopsies when applied as rule-out tests in elderly patients.

Follicular patterned nodules are sometimes complex to be classified due to ambiguous nuclear features and/or questionable capsular or vascular invasion. In this setting, there is a poor inter-observer concordance even among expert pathologists. Raman spectroscopy was recently used to separate benign and malignant thyroid nodules based on their molecular fingerprint, anyway, some histologically proved follicular adenomas were clustered as having a characteristic profile of malignant lesions. In this study, we analyzed five follicular thyroid adenomas with a malignant spectroscopic profile compared to five follicular adenomas with a benign Raman spectrum in order to assess possible molecular differences between the two groups. Morphological, immunohistochemical, and molecular analyses evidenced expression of malignancy-associated proteins in four out of five malignant clustered adenomas. The remaining malignant clustered adenoma showed a TSHR mutation previously associated with autonomously functioning follicular carcinomas. In conclusion, thyroid follicular adenomas are a group of morphologically benign neoplasms that may have altered the mutational or expression profile, cases of adenomas with altered immunophenotype are recognized as showing a profile associated with malignancy by Raman spectroscopy. This correlation warrants a more extensive evaluation and suggests a potential predictive value of spectroscopic assessment in recognizing characteristics associated with tumor progression in follicular thyroid neoplasms.

H&uuml, rthle cell carcinomas (HCC) are rare differentiated thyroid cancers that display low avidity for radioactive iodine and respond poorly to kinase inhibitors. Here, using next-generation sequencing, we analyzed the mutational status of primary tissue and poorly differentiated metastatic tissue from two HCC patients. In both cases, metastatic tissues harbored a mutation of SETD2, each resulting in loss of the SRI and WW domains of SETD2, a methyltransferase that trimethylates H3K36 (H3K36me3) and also interacts with p53 to promote its stability. Functional studies of the novel p.D1890fs6* mutation (case 1) revealed significantly reduced H3K36me3 levels in SETD2-mutated tissue and primary cell cultures and decreased levels of the active form of p53. Restoration of SETD2-wildtype expression in the SETD2-mutant cells significantly reduced the expression of four well-known stemness markers (OCT-4, SOX2, IPF1, Goosecoid). These findings suggest potential roles for SETD2 loss-of-function mutations in HCC progression, possibly involving p53 destabilization and promotion of stemness. Their prevalence and potential treatment implications in thyroid cancer, especially HCC, require further study.

It is unclear whether the site of origin of papillary thyroid microcarcinoma (mPTC) with respect to the thyroid surface has an influence on clinicopathologic parameters. The objectives of the study were to: (i) Accurately measure the mPTC distance from the thyroid surface, (ii) analyze whether this distance correlates with relevant clinicopathologic parameters, and (iii) investigate the impact of the site of origin of the mPTC on risk stratification. Clinicopathologic features and BRAF mutational status were analyzed and correlated with the site of origin of the mPTC in a multicenter cohort of 298 mPTCs from six Italian medical institutions. Tumors arise at a median distance of 3.5 mm below the surface of the thyroid gland. Statistical analysis identified four distinct clusters. Group A, mPTC: size &ge, 5 mm and distance of the edge of the tumor from the thyroid capsule = 0 mm, group B, mPTC: size &ge, 5 mm and distance of the edge of the tumor from the thyroid capsule &gt, 0 mm, group C, mPTC: size &lt, and group D, mPTC: size &lt, 0 mm. Univariate analysis demonstrates significant differences between the groups: Group A shows the most aggressive features, and group D the most indolent ones. By multivariate analysis, group A tumors are characterized by tall cell histotype, BRAF V600E mutation, tumor fibrosis, aggressive growth with invasive features, vascular invasion, lymph node metastases, and intermediate ATA risk. The mPTC clinicopathologic features vary according to the tumor size and distance from the thyroid surface. A four-group model may be useful for risk stratification and to refine the selection of nodules to be targeted for fine needle aspiration.