Malignancy-risk stratification of thyroid nodules : diagnostic utility of the ultrasound computer-aided diagnosis and advanced ultrasound imaging techniques

Abstract

Thyroid cancer is the most prevalent endocrine cancer which typically manifests as thyroid nodules that are primarily diagnosed using ultrasound and fine-needle aspiration cytology (FNAC). The surge in thyroid cancer incidence and the overdiagnosis of indolent thyroid nodules in recent years is attributed to the increased use of advanced ultrasound technology and FNAC. This thesis evaluated the diagnostic utility of computer-aided diagnosis (CAD), AngioPLUS microvascularity assessment and shear wave elastography (SWE) in thyroid cancer diagnosis. Study 1 validated the diagnostic performance of CAD software based on 6 TIRADS using 205 thyroid nodules at the default setting and 3 adjusted settings. The results confirmed that the default setting was ideal for achieving maximum sensitivity. The EU TIRADS demonstrated the highest sensitivity (82.7%). Comparative diagnostic performance analysis was done between the CAD and computer-assisted subjective interpretation using 4 analogous TIRADS and 162 nodules. The results showed comparable sensitivity between the two methods, however, CAD had lower specificity for all TIRADS. EU TIRADS and KSThR TIRADS differentiated all nodules with comparable diagnostic performance, however, EU TIRADS had slightly higher specificity. Study 2 evaluated the diagnostic value of AngioPLUS coupled with conventional power (APDI) and colour Doppler (ACDI) ultrasound imaging techniques in combination with EU TIRADS. A total of 94 thyroid nodules that included 40 cytologically-equivocal nodules were analysed using a quantitative ratio vascularity index (RVI) at different offsets and qualitative visual regional grading. RVI combined with EU TIRADS resulted in high specificity but poor sensitivity at all offsets and Doppler ultrasound modes. The combination of qualitative APDI vascularity grading and EU TIRADS had higher specificity than EU TIRADS alone and slightly lower sensitivity in differentiating all nodules (SEN: 76.7% vs 83.3%, p > 0.05, SPEC: 84.4% vs 50%, p < 0.05;). However, it stratified cytologically-equivocal nodules with significantly higher overall diagnostic performance than EU TIRADS alone (SEN: 88.9%; SPEC: 77.4% vs 38.7%, p < 0.05, and AUROC: 0.89 vs 0.62, p < 0.05). Study 3 evaluated the diagnostic value of SWE in combination with EU TIRADS and included 126 thyroid nodules. Using different sub-categories of the nodules, diagnostic performance analyses were performed. Results showed that no SWE index was superior to EU TIRADS in differentiating all nodules. However, for nodules between 1cm and 2 cm at an optimal cut-off of 8.7 kPa and those > 2cm at an optimal cut-off of 10.7 kPa, the SWE SD index combined with EU TIRADS demonstrated higher specificity and insignificantly lower sensitivity than EU TIRADS alone (SEN:72.2% vs 88.9%, p > 0.05; SPEC:76.5% vs 55.9%, p < 0.05; and, SEN: 71.4% vs 85.7%, p > 0.05; SPEC: 95.8% vs 62.5%, p < 0.05, respectively). In conclusion, CAD has a complementary diagnosis role in thyroid cancer screening, but it is not superior to subjective interpretation. EU TIRADS can rule in thyroid malignancy using either CAD or subjective interpretation. APDI combined with EU TIRADS is highly diagnostically efficient in differentiating cytologically-equivocal thyroid nodules. SWE combined with EU TIRADS has the potential for improving the diagnosis of nodules ≥ 1cm.