51 Diagnostic utility of molecular and imaging biomarkers 2 wide shape was generally specific for carcinoma, with reported specificities up to 99% [239, 240, 242]. A spherical shape is generally considered benign, but has also been associated with FTC [229, 231, 263]. In two studies in cytological follicular neoplasms, a spherical shape had an increased risk of malignancy, with 86% to 97% sensitivity and 19% to 26% specificity [264, 265]. Chin et al. even suggested that follicular neoplasms with a taller-than-wide shape could be treated conservatively [265]. The uniquely balanced rates of PTC, FVPTC and FTC resulting from indeterminate cytology may explain why these and various other US characteristics have different diagnostic accuracy than in the unselected population. Dependent on the local case mix, accurate differentiation of indeterminate nodules using the classical suspicious US features may or may not be feasible. Combination of ultrasound characteristics A combination of US characteristics likely provides more accurate differentiation than individual features. Different combinations were investigated in multiple studies [119, 235, 236, 238, 242-244, 254, 266-271]. Yoo et al. reported 100% specificity for the combination of marked hypoechogenicity and taller-than-wide shape, a pattern that occurred in 9.6% (24/249) of the included Bethesda III nodules [239]. In the elastosonography study by Rago et al., absence of a hypoechoic halo in combination with presence of microcalcifications was 95% specific for thyroid malignancy, but only 6.4% sensitive [260]. Maia et al. found 62% sensitivity and 89% specificity in Bethesda III and IV nodules if hypoechogenicity, microcalcifications, an irregular margin and increased intranodular vascularity were considered suspicious [268]. Gulcelik et al. demonstrated that the US pattern of a solid, hypoechoic nodule with microcalcifications had 95% sensitivity and 99% specificity. The pattern was seen in 21% of cytological follicular neoplasms [272]. In multiple studies it was argued that cytological follicular neoplasms with a typically benign ultrasound pattern – a regular shape, isoechoic, homogeneous, with well-defined margins, cystic components or peripheral vascularity only, and not a single malignant feature - could be safely followed up clinically instead of undergoing diagnostic surgery [238, 263, 266]. Consideration of more features generally increased the sensitivity of the US assessment at the cost of its specificity [236, 237]. The terms of their interpretation were crucial: Norlén et al. demonstrated that US was 95% sensitive and 48% specific if a Bethesda III nodule had either hypoechoic appearance, irregular margins or microcalcifications. If solely the simultaneous presence of all three features was considered suspicious for malignancy, sensitivity dropped to 37% but specificity increased to 96% [269]. Altogether, diagnostic ultrasound scores or step-by-step algorithms could aid the classification of US patterns and consequent risk of malignancy [238, 266, 273]. Best-known and most validated is the Thyroid Imaging Reporting and Data System (TIRADS), a classification to risk-stratify thyroid nodules, designed by Horvath et al. and modified by Kwak et al. following the example of the similar BIRADS classification for breast lesions [274, 275]. The TIRADS assigns nodules to a risk category based on five suspicious US features: solid appearance, (marked) hypoechogenicity, irregular margins, microcalcifications and a taller-than-wide shape. Nodules without any of these features
RkJQdWJsaXNoZXIy MTk4NDMw