Ultrasonography (US) is the main imaging method used to evaluate thyroid nodules, and it plays a central role in detection and characterization. It not only helps distinguish malignant from benign lesions but also guides fine‑needle aspiration (FNA) when malignancy is suspected [1,2]. Color Doppler imaging (CDI) has been widely used to assess vascular patterns of thyroid nodules, contributing to diagnostic evaluation. Multiple guidelines emphasize that US features and risk stratification systems (RSSs) are useful for estimating malignancy risk in thyroid nodules [3–7]. Although numerous RSSs list US features suggestive of malignancy, there is no consensus on the diagnostic significance of nodular vascularity. The rationale for assessing vascularity derives from the observation that proliferative activity may increase vascularization, which can resemble aberrant angiogenesis associated with malignant transformation [8,9]. Nevertheless, whether thyroid cancers actually show increased Doppler vascularity remains a matter of debate [10]. Because the proportions of pathological cancer subtypes differ among studies, findings regarding nodular vascularity have been inconsistent. The American Thyroid Association guidelines suggest that variability in study outcomes may depend on the proportion of follicular thyroid carcinomas (FTCs) [3-7], since FTCs are highly vascular and tend to metastasize hematogenously, whereas papillary thyroid carcinomas (PTCs) are rich in lymphatic vessels and typically metastasize to lymph nodes [11].

Recently, several US manufacturers have introduced techniques for visualizing blood flow in microvessels, collectively termed microvascular flow imaging (MVFI) [12]. This technology adaptively filters random motion while preserving the directional flow of blood, enabling more sensitive and higher-resolution assessments than conventional color or power Doppler US [12]. In other malignancies, such as hepatocellular carcinoma, tumor angiogenesis and neovascularization make CDI and microflow imaging (MFI)—a specific MVFI technique developed by Philips Healthcare (Eindhoven, Netherlands)—valuable diagnostic tools [13]. However, debates persist regarding whether color flow signals are helpful in differentiating malignant from benign thyroid nodules. Consequently, vascular findings have not been incorporated as consensus diagnostic criteria in various RSSs.

Therefore, this study aimed to determine whether vascular scores obtained using CDI and MFI are useful in distinguishing malignant from benign thyroid nodules and to evaluate the correlation between these scores and the Thyroid Imaging Reporting and Data System (TI-RADS) classification. Additionally, by examining the relationship between vascular scores and the pathological subtypes of malignancy, this study sought to clarify the potential diagnostic role of vascularity in thyroid cancer.

Among the 113 thyroid nodules analyzed, 72 were benign and 41 were malignant. The lesion diameters on B-mode US ranged from 0.5 cm to 5.6 cm (median, 1.2 cm). Surgical confirmation was obtained for 33 nodules. Among these, malignant lesions consisted of 20 classic PTCs and eight follicular variants of PTC, while benign lesions included four nodular hyperplasias and one follicular adenoma (Table 1). The incidence of DTD was comparable between malignant and benign groups (6.9% [5/72] vs. 7.3% [3/41], respectively) and showed no significant difference between K-TIRADS categories 3/4 and category 5 nodules (7.3% [6/82] vs. 6.7% [2/30], respectively).

Table 2 and Supplementary Table 1 summarize the ultrasonographic findings of thyroid nodules according to the K-TIRADS lexicon. All lexicon features—including composition, echogenicity, orientation, margin, and calcifications—significantly differentiated malignant from benign lesions (P≤0.008) (Supplementary Table 1). The nodules were classified by K-TIRADS category as follows: category 2 (n=1, 0.9%), category 3 (n=39, 34.5%), category 4 (n=43, 38.1%), and category 5 (n=30, 26.5%). Regarding vascular scores, the CDI score differed significantly between malignant and benign nodules (P<0.001), whereas the MFI score did not show a significant difference (P=0.174). Benign nodules more frequently exhibited higher vascular scores (CDI or MFI scores 3–4: 52.8% and 70.8%, respectively) compared with malignant nodules (CDI or MFI scores 3–4: 17.1% and 51.2%, respectively), while malignant nodules more often demonstrated lower vascular scores (Table 2). For the vascular score, both intra-observer and inter-observer variability demonstrated moderate to substantial agreement in CDI and MFI for both reviewers. The intra-observer agreement was κ=0.77 for reviewer 1 and κ=0.79 for reviewer 2 in CDI, and κ=0.80 for reviewer 1 and κ=0.79 for reviewer 2 in MFI. The inter-observer agreement was κ=0.60 for CDI and κ=0.59 for MFI.

To further evaluate vascularity, a subgroup analysis was performed according to K-TIRADS categories (Table 3). The vascular scores did not differ significantly between malignant and benign nodules in K-TIRADS categories 3 and 4 (P>0.05). However, in K-TIRADS 5 nodules, vascular scores differed significantly between malignant and benign lesions (P=0.018 for CDI and P=0.051 for MFI). The vascular scores were relatively higher in benign nodules than in malignant ones (Fig. 2).

Table 4 presents a comparison of vascular scores according to K-TIRADS categories. K-TIRADS 3 and 4 nodules showed similar vascularity regardless of benign or malignant status. However, malignant nodules classified as K-TIRADS 5 tended to exhibit lower vascularity than benign nodules, in both CDI and MFI (P=0.012 and P=0.070, respectively) (Fig. 3). In the subgroup analysis of small nodules, K-TIRADS 5 lesions demonstrated similar trends, although the differences did not reach statistical significance (P=0.091 for both CDI and MFI) (Supplementary Table 2). In the analysis of vascular scores based on pathologic subtypes of malignancy, classic PTC showed lower vascular scores in both CDI and MFI, whereas the follicular variant of PTC exhibited higher vascular scores (P=0.075 and P=0.007, respectively). Notably, no vascular score of 1 was observed in the follicular variant of PTC, and no score of 4 was recorded in classic PTC on either CDI or MFI. Despite these trends, the pathologic subtypes of malignancy did not show a significant correlation with K-TIRADS categories (P=0.406) (Supplementary Table 3).

This study investigated whether vascular scoring could predict thyroid cancer and explored its potential integration into the TI-RADS system to refine malignancy risk assessment. When all nodules were analyzed collectively, benign lesions exhibited higher CDI scores than malignant ones. In contrast, within K-TIRADS categories 3 and 4, there was no meaningful difference in vascularity between malignant and benign nodules. Among K-TIRADS 5 nodules, benign lesions showed higher vascular scores, whereas malignant lesions demonstrated relatively lower values. Within the malignant group, the follicular variant of PTC tended to display higher vascularity than classic PTC. However, the number of nodules in certain subgroups—particularly K-TIRADS 5—was small, limiting statistical power and generalizability. To assess the potential impact of background parenchymal vascularity, we also examined the prevalence of DTD, which can lead to either an increase or a decrease in thyroid gland vascularity, and variations in disease severity may influence the observed vascularity measurements [16]. In this study, the prevalence of DTD did not differ significantly between malignant and benign nodules or across K-TIRADS categories, suggesting that DTD likely did not substantially affect the observed differences in vascularity between groups.

Representative neoplastic vascular alterations in malignancies include both morphological and hemodynamic changes, typically characterized by hypervascularity and the presence of arteriovenous shunting [9]. However, internal vascularization in thyroid nodules can arise not only from neoplastic vascular proliferation but also from hyperplastic follicular proliferation or granulation tissue in nodules of colloid goiter [17,18]. Therefore, uncertainty persists regarding whether vascularity can reliably differentiate malignant from benign thyroid nodules. Yang et al. suggested that inconsistent findings regarding thyroid nodule vascularity across studies may reflect varying proportions of pathological cancer subtypes, as FTCs and PTCs differ microscopically [10]. PTCs show dense fibrosis in 56%–89% of cases [19–21], whereas FTCs are characterized by abundant blood vessels [11]. The authors also proposed that most hypervascular thyroid nodules correspond to adenomatoid nodules, the encapsulated form of the follicular variant of PTC, or to follicular carcinomas; however, malignant nodules generally lack intranodular vascularity, consistent with the present findings [10]. Moon et al. [22] similarly reported that benign thyroid nodules more frequently exhibit intranodular vascularity, while malignant nodules are more often avascular. Using power Doppler US, they concluded that vascularity was less useful than grayscale US features in predicting thyroid malignancy. Unlike earlier investigations [10,22], the present study assessed nodular vascularity within the structured framework of the TI-RADS system, which uses a lexicon-based approach to stratify malignancy risk. PTC and FTC are known to present distinct ultrasonographic features, and these differences may aid in predicting histologic subtypes of thyroid cancer [23,24]. A prior study analyzing malignant tumor types across K-TIRADS categories reported that PTC was the predominant malignancy among K-TIRADS category 5 nodules, while FTC predominated in categories 3 and 4 [25]. PTC typically manifests as K-TIRADS 5 nodules and is often associated with dense fibrosis; therefore, vascularity assessment may be more useful in differentiating malignant from benign nodules in K-TIRADS 5 than in K-TIRADS 3 or 4. Taken together, these findings and the prior literature suggest that the diagnostic utility of vascularity in thyroid cancer should be interpreted in the context of the TI-RADS category.

In this study, 32 cases of classic PTC and eight cases of the follicular variant of PTC were included among the malignant lesions. Classic PTC generally showed lower vascularity than the follicular variant in both CDI and MFI. Although the follicular variant of PTC belongs to the broader PTC group, it is histologically characterized by a predominantly follicular growth pattern [26]. In contrast, classic PTC displays papillary architecture with central fibrovascular cores. Lamellated calcific spherules (psammoma bodies) are observed in approximately 40%–60% of cases, and degenerative changes with cystic formation are frequent. The stroma in classic PTC is often desmoplastic and may be extensive [26]. These histopathologic distinctions between the two PTC subtypes may underlie the vascularity differences observed in the present study. Although no significant difference in K-TIRADS categories was identified between the two PTC variants, further research with larger and more histologically diverse populations is warranted to clarify these associations.

The MVFI technique enables visualization of small vessels and detection of slow blood flow without using contrast agents [12,13]. Its main advantages include the ability to depict low-velocity blood flow clearly, provide high spatial resolution, reduce motion-related artifacts, and maintain high frame rates. Since its clinical introduction, numerous studies have explored the potential applications of MVFI across various organs and disease conditions, including the thyroid gland [12]. Some studies have reported that MVFI outperforms color or power Doppler imaging in evaluating malignant thyroid nodules, with findings such as abundant vascularity and intranodular flow associated with malignancy [27–29]. However, a definitive consensus on the diagnostic utility of MVFI in thyroid nodules has yet to be reached. Therefore, in this study, nodular vascularity assessed by MVFI was applied to the TI-RADS classification system, representing, to the authors’ knowledge, the first attempt to integrate MVFI-derived vascular data into TI-RADS. Although the results did not show a statistically significant difference between MVFI and CDI, they support the notion that vascularity should be interpreted differently depending on the K-TIRADS category. Future studies with larger sample sizes and broader malignancy spectra are necessary to further define the clinical role and diagnostic utility of MVFI in thyroid nodule assessment.

This study has several limitations. First, it was limited by its relatively small sample size and single-center, retrospective design, which may have introduced selection bias, including restricted pathological diversity. The proportion of malignant nodules was relatively high, and these nodules tended to be smaller in size, likely reflecting the tertiary referral setting. Moreover, the sample size was insufficient to robustly evaluate differences in MFI scores across K-TIRADS categories. In particular, the findings for K-TIRADS 5 were based on only four benign nodules, rendering these results preliminary and limiting both their statistical reliability and generalizability. Future multicenter studies with larger cohorts are warranted to validate these findings. Second, US and Doppler imaging are inherently subject to substantial inter-operator and inter-equipment variability. In this study, all CDI and MFI examinations were performed by a single experienced radiologist using the same US system to minimize variability. However, this approach may also limit the generalizability of the results. In addition, vascularity was evaluated using a four-point scale by two reviewers through consensus, which precluded assessment of inter- and intra-observer variability in vascularity scoring. Third, all US examinations were performed using equipment from a single manufacturer, which may limit the applicability of these findings to other US systems. Future validation studies involving multiple operators and diverse equipment are therefore needed. Finally, 13 cases (approximately 32% of malignancies) were diagnosed by FNA or CNB without surgical confirmation, 12 of which were classified as classic PTC. Given the inherent difficulty in differentiating classic PTC from the follicular variant of PTC (FVPTC) using FNA alone, these slides were re-reviewed by an experienced pathologist, confirming consistency with classic PTC. Nonetheless, the possibility of other PTC subtypes, including FVPTC, cannot be completely excluded. Furthermore, the relatively small total sample size (n=113) and limited number of FVPTCs (n=8) may restrict the generalizability of the findings. Despite these limitations, this study provides important preliminary insights into the potential role of CDI and MFI in distinguishing malignant from benign thyroid nodules. Larger-scale studies are necessary to confirm these results and to further evaluate US vascularity across different pathological subtypes of thyroid cancer.

In conclusion, vascular scores obtained from CDI and MFI appear to be valuable US parameters for differentiating malignant from benign thyroid nodules within the TI-RADS classification framework. The diagnostic application of nodular vascularity in thyroid cancer should be tailored according to the TI-RADS category, which may reflect the underlying pathological subtype of malignancy.