In this study, we demonstrated that SMI can indicate characteristic blood flow distribution of metastatic LNs with PTC, improving the preoperative diagnostic accuracy of metastatic LNs. Compared with US alone, SMI combined with US improved accuracy, specificity, and PPV, with AUC of 0.829 and 0.926, respectively.
Vascularity was considered an important characteristic in metastatic cervical LNs with PTC, which was confirmed by pathological evidence in a recent study showing that microvessel density and microvessel area in the metastatic LN group were significantly higher than those in the non-metastatic LN group [Zhining et al.2022]. Similarly, in this study, the SMI vascularity index between level 2–3 in metastatic LNs (56%) was significantly higher than that in reactive LNs (23%) (Table 1). Therefore, we suppose that compared to US, SMI can optimize the diagnosis of metastatic LNs.
Angiogenesis is essential for tumorigenesis, proliferation and invasion [Folkman et al.1971]. The particular vascular distribution in malignant LNs is related to neovascularization, desmoplastic reaction, and capsule vessel proliferation. Tumor cells enter the LNs through afferent lymphatic vessels, cell proliferation leads to cortex enlargement, and tumor cells invade the remaining normal tissue [Bonacchi et al.2017]. Hence, peripheral and mixed vascular distribution was present in most malignant LNs, which was consistent with the results of Sim et al [Sim er al.2019; Ryoo et al.2016; Seongyong et al.2020], and the hilar vascular pattern was the most common in normal or reactive LNs [Ahuja et al.2003]. Similarly, in this study, most metastatic LNs exhibited a peripheral mixed vascular pattern (94%) and reactive LNs mainly exhibited avascular or hilar vascular pattern (79%). Notably, we found several special vascular characteristics of SMI emerging in partially metastatic LNs but rarely seen in reactive LNs, as mentioned above (see Result 5. Special vascular characteristics of SMI). Consistently with this, our previous studies reported that metastatic LNs manifested heterogeneous enhancement and ring-enhancing margins more often than benign LNs on preoperative CEUS [YR Hong et al.2017; ZY Luo et al.2022]. We believe that this special vascular distribution is related to the pathological mechanism of tumor formation. The ring-shaped vascular distribution is likely due to metastatic cells entering LNs through afferent lymphatic vessels and spreading from the marginal sinuses [Ji N et al.2020; Fengkai et al.2022]. The focally abundant and misty flow performance are due to rapid cell proliferation and immature neovascularization [Fengkai et al.2022]. In addition, the indolent characteristic of PTC result in incomplete invasion of cervical LNs by tumor cell, which also contributes to focal abundance and misty flow performance.
With the emergence of SMI in 2014, based on the Canon medical system [Ohno et al.2017], many studies have proven its ability to diagnose human diseases, such as tumors, inflammation and injury, owing to visualization of small vessels with slow flow in tissues [Ahuja et al.2003; Bae et al.2020; Sato et al.2021; De Backer et al.2013]. CEUS is a promising tool for providing capillary vessel perfusion information [YR Hong et al.2017; Fengkai et al.2022]. A previous study proved that the diagnostic value of SMI was comparable with CEUS in thyroid nodules [YF Zhao et al.2020].
Furthermore, a nomogram based on the training dataset was conducted to evaluate the predictive value of SMI and US for metastatic LNs with PTC. Parameters with statistical significance in the univariate analysis were selected for multivariate analysis. The results revealed that L/S < 2 and peripheral or mixed vascular type on CDFI and peripheral or mixed vascular type on SMI were independent predictors of metastatic LNs in patients with PTC. Peripheral or mixed vascular type on SMI had the highest predictive value with the highest odds ratio (OR = 40.829), compared to L/S < 2 (OR = 8.684) and mixed vascular type on CDFI (OR = 5.982).
This study had several limitations. First, a potential selection bias is inevitable in this retrospective study. SMI was not a routine examination in the clinic practice, and partially reactive LNs which underwent SMI were considered suspicious on US. Second, the patients were enrolled at a single center; observer bias was present in this study, and a multicenter study with a larger number of patients is required. Third, the imaging findings of the vascular structures did not correlate with the true specimen pathology. Further research on pathological vessel distribution corresponding to SMI imaging is needed.
In conclusion, our findings demonstrate that SMI can significantly improve the diagnostic accuracy of metastatic cervical LNs from PTC. We also established and validated an accurate nomogram for predicting the possibility of LN metastasis in patients with PTC. Preoperative SMI has the significant potential to enhance the ability of surgeons to determine optimal clinical management.