Ultrasound features of thyroid follicular adenoma and follicular carcinoma: A narrative review.

Introduction and research background

Epidemiology and clinical significance of thyroid follicular tumor
As an important subtype of thyroid epithelial tumors, thyroid follicular tumors involve special diagnostic challenges in clinical practice. Epidemiological data have shown that follicular thyroid adenomas (FTAs) account for 15%–20% of all benign thyroid tumors, while follicular thyroid carcinomas (FTCs) account for 10%–15% of all malignant thyroid tumors, representing the second most common type of carcinoma after papillary carcinoma.
1
Notably, the overall incidence of thyroid cancer has shown a significant upward trend in recent years, with an increase of more than three times over the past three decades.
2
The clinical importance of this type of tumor is reflected in its diagnostic dilemma; FTA and FTC are almost identical in terms of cell morphology and can only be distinguished on the basis of histological features such as capsule or vascular invasion.1,3 Owing to this biological characteristic, the rate of accurate preoperative diagnosis rate is <50%, resulting in a large number of patients with benign nodules undergoing unnecessary surgical treatment.
4

The gold standard of pathological diagnosis and the clinical need for preoperative evaluation
Histopathological examination is the only gold standard for distinguishing an FTA from FTC. It is necessary to comprehensively evaluate tumor capsule integrity and vascular invasion.3,5 However, this diagnostic model has obvious clinical limitations, and all patients with follicular tumors must undergo diagnostic surgery for clear pathological conclusions.
6
A large-scale multicenter study showed that in 3634 cases of follicular tumors confirmed using pathological evaluations, benign adenomas accounted for 48.1% of cases (1748 cases), while malignant cases accounted for only 8.2% (299 cases).
4
This results in considerable wastage of medical resources and increased patient burden. The new classification of noninvasive follicular thyroid tumor (NIFTP) introduced by the World Health Organization (WHO) in 2017 has further complicated the situation by expanding the pathological spectrum of follicular tumors, thereby increasing the urgency of preoperative differential diagnosis.
7

The core value of ultrasound in differential diagnosis
Owing to its advantages such as noninvasiveness, real-time evaluation, and low cost, ultrasound has become a first-line imaging tool for the evaluation of thyroid nodules.2,7 Existing studies have shown a potential correlation between ultrasound characteristics and tumor biological behavior, such as the taller-than-wide (TTW) sign and other morphological parameters, showing differential distribution in benign and malignant follicular tumors.
8
However, traditional ultrasound risk stratification systems, such as the thyroid imaging reporting and data system (TI-RADS), are mainly designed for evaluating papillary cancer, and their differential ability to classify follicular tumors is limited.6,9 A recent study analyzed 9393 ultrasound images of follicular tumors and showed that the deep learning (DL) system could identify malignant tumors with 78.3% accuracy.
10
New modalities such as contrast-enhanced ultrasound (CEUS) and elastography can provide supplementary information about tumor microcirculation and hardness, thereby demonstrating their potential in overcoming existing diagnostic challenges.6,11 These technological advancements have enabled the evolution of ultrasound from a simple morphological observation tool to a multiparameter function evaluation system.2,12

Methods
This narrative review was conducted according to the Scale for the Assessment of Narrative Review Articles (SANRA) guidelines
13
to ensure a structured and comprehensive synthesis of the literature. Although this was not a systematic review, a targeted literature search was performed in PubMed and Google Scholar for articles published between 2000 and 2025. Search terms included combinations of “thyroid follicular adenoma,” “follicular thyroid carcinoma,” “ultrasonography,” “elastography,” “contrast-enhanced ultrasound,” and “artificial intelligence.” The scope was limited to English-language articles. The selection of studies was based on their relevance to the aim of our review, which was to summarize and compare the ultrasonographic features of FTAs and FTCs, discuss diagnostic controversies, and evaluate emerging technologies. Given the narrative nature of this review, formal study quality assessment or statistical pooling of data was not performed.

Systematic analyses of the ultrasonographic characteristics of FTAs

Typical ultrasonographic manifestations: shape, margin, and vascularity characteristics
FTAs usually appear as oval or round nodules with clear margins and regular shapes on ultrasound examination.
14
Multicenter research data have shown that approximately 78% of FTAs present with complete hypoechoic halo, which is highly consistent with the pathological evidence of tumor capsule integrity.
15
In terms of echogenicity characteristics, FTA mostly shows uniform isoechogenicity or slightly hyperechogenicity, with a clear margin from the surrounding normal thyroid tissue.
2
Vascularity pattern analysis showed that FTAs mostly exhibit peripheral or mixed vascularity signals, of which peripheral vascularity signals account for 62.3%; furthermore, the intensity of central vascularity signals is usually lower than that of the peripheral regions.
16
Notably, approximately 15%–20% of FTAs may show ultrasound features similar to those of malignant nodules, such as microcalcification and vascularity disorder, which reduces the diagnostic specificity.
17

Ultrasonographic variation characteristics of special subtypes
FTAs of special histological subtypes show unique ultrasonographic variations. Follicular adenoma with papillary structure often presents with cystic and solid mixed echogenicity on ultrasound, and incomplete separation is observed in approximately 34% of cases.
18
Oncocytic adenoma is characterized by significant hypoechogenicity and posterior echogenicity enhancement, with an internal echogenicity nonuniformity rate of 89%.
19
A large-sample study (n = 1748) found that FTAs with hyaline degeneration showed a characteristic “star shaped” central scar on ultrasound, which was particularly obvious on coronal section.
2
In addition, in approximately 8.7% of FTA cases, the phenomenon of “hyperechoic spots” can be observed, which has been confirmed on histology to correspond to the glial concentration area.
19
These varied characteristics increase the difficulty of distinguishing an FTA from FTC, especially in cases lacking typical malignant signs.
20

Study on the correlation analysis between ultrasonography and histopathology
The correlation analysis between ultrasonographic characteristics and FTA histopathology showed a significant correlation between the ultrasonographic evaluation of nodule margin and histological capsule integrity (κ = 0.72).
6
The minimum capsule thickness detected using high-resolution ultrasound was 0.3 mm, which was highly consistent with the pathological results (r = 0.81, p < 0.001).
21
In terms of internal structure, the “spongy” echogenicity mode displayed on ultrasound corresponded to the aggregation of the microfibrillar structure in histology, and its positive predictive value reached 91%.
12
Quantitative analysis of vascularity characteristics and vascular density showed that the microvessel density (MVD) of peripheral vascularity–dominant FTA was significantly lower than that of the central vascularity type (p = 0.003).
22
A recent study that performed ultrasound omics analysis reported that the model based on texture features can predict the molecular subtypes of FTA; the identification accuracy of RAS mutant adenoma was up to 82.5% in this study.
23
These findings provide a new research direction for the preoperative noninvasive evaluation of the biological characteristics of FTA.
24

Systematic analysis of the ultrasonographic characteristics of FTC

Ultrasonographic markers of invasive characteristics (e.g. capsule invasion/vascular invasion)
The most valuable ultrasonographic feature of FTC is its invasiveness. Capsule invasion can manifest as an irregular nodule margin, interrupted capsule, or convex growth on ultrasound, which are significantly correlated with capsule penetration on pathology.
5
Ultrasonographic markers of vascular invasion include abnormal thickened vascular structure (diameter >0.5 mm) inside the nodule and distortion of peripheral blood vessels, and color Doppler imaging shows disordered vascularity signals.5,19 Notably, the 2022 WHO classification defines invasive nonanaplastic thyroid cancer with high-grade characteristics as high-grade follicular cell-derived thyroid cancer (HGFCTC), which is more likely to show a thick and irregular calcified ring with posterior acoustic shadow on ultrasound.16,23 For invasive follicular variant-papillary thyroid carcinoma (IFV-PTC), its invasive growth pattern can be characterized by a “crab foot like” marginal invasion of the surrounding thyroid parenchyma on ultrasound.
22

Ultrasonographic spectrum of different histological grades
According to the WHO classification system, the ultrasonographic manifestations of FTC exhibit significant differences in histological grading. Well-differentiated FTC usually presents as isoechoic or hypoechoic nodules with relatively clear margins, which may be accompanied with thin and complete halo rings.16,25 Poorly differentiated thyroid cancer (PDTC) and well-differentiated high-grade thyroid cancer (HGdTC) tend to show the following characteristics: maximum nodule diameter >4 cm, internal coarse calcification with acoustic shadow, and diffuse vascularity signal.16,26 It is particularly noteworthy that FTC with high-grade features shows higher tissue hardness (average Young’s modulus >65 kPa) on ultrasound elastography.
25
Molecular characteristics analysis has shown that follicular tumors with RAS mutation are more likely to show a uniform echo structure on ultrasound, while cases with high-grade histology and TP53 or TERT promoter mutation exhibit mixed echogenicity.19,26

Differential characteristics of microinvasive-type and extensive invasive–type FTC
Microinvasive FTC (only microscopic capsule or vascular invasion) is often difficult to distinguish from benign adenoma using conventional ultrasound. Most of them are nodules with clear margins and uniform echogenicity. Only approximately 28% of cases show an incomplete halo.5,26 In contrast, extensively infiltrating FTC has more typical malignant characteristics such as an unclear margin between the nodule and surrounding tissue, internal microcalcification (diameter <1 mm), and Doppler mode dominated by central vascularity.4,5,20 Regarding NIFTP characteristics, its ultrasound features are more similar to those of benign lesions, usually presenting with a round or oval shape, with well-defined noncalcified nodules, which is in sharp contrast to the features of invasive encapsulated follicular variant papillary carcinoma (IENC-FVPTC).6,27 Elastic imaging parameters can provide additional differential value. The strain ratio of extensive infiltrating FTC is usually >4.5, while that of microinvasive FTC is <3.0 in most cases.4,24

Comparative study on the ultrasonographic characteristics of FTA and FTC

Quantitative comparison of morphological parameters (e.g. aspect ratio and margin regularity)
There were significant differences in the morphological parameters between FTA and FTC. Studies have shown that FTC nodules are more often characterized by the TTW sign, which holds value in differential diagnosis.
28
Quantitative analyses have demonstrated that the average aspect ratio of FTC is significantly higher than that of FTA, and the nodule sphericity index is significantly different between the two groups.
28
In terms of margin features, FTC is more prone to irregular boundaries, while FTA is usually characterized by clear and regular margins.
29
A multicenter, retrospective analysis involving 3634 patients has demonstrated good identification efficiency of these morphological parameters.
26

Differences between internal echogenicity characteristics and calcification patterns
There are significant differences in the internal echogenicity characteristics between FTA and FTC. Research has shown that FTC tends to show uneven low echogenicity, while FTA shows uniform isoechogenicity or slightly high echogenicity.
30
The incidence of microcalcification in FTC is significantly higher than that in FTA, which is correlated with pathological nuclear atypia.
29
Notably, analysis of 9393 ultrasound images has shown that the occurrence rate of coarse calcification in the FTC group was higher than that in the FTA group; however, the statistical significance of this difference remains controversial.
31
In addition, special subtypes such as FVPTC and NIFTP show overlap with typical FTA in terms of echogenicity characteristics, which increases the difficulty of differential diagnosis.7,32

Controversy regarding the diagnostic threshold of hemodynamic parameters
There is considerable debate regarding the hemodynamic parameters used for distinguishing FTAs from FTCs. Although many studies have confirmed that the vascularity signal intensity of FTC is usually higher than that of FTA, there is no consensus on the specific diagnostic threshold.
29
CEUS studies have shown that FTC is characterized by rapid enhancement in the early stage and rapid regression in the late stage, while FTAs exhibit slow enhancement; however, there is significant overlap in the enhancement intensity.
33
Doppler parameter analyses have shown that the average values of resistance index (RI) and pulsatility index (PI) for FTC were higher than those for FTA; however, the optimal diagnostic threshold values were significantly different in distinct studies (RI: 0.65–0.75, PI: 1.1–1.3).32,33 This difference can be attributed to the heterogeneity of equipment parameters and measurement methods used in different studies, suggesting the need to establish a standardized vascularity assessment scheme (Table 1).26,31

Technical disputes over current diagnostic criteria

Limitations of TI-RADS classification in follicular tumors
The current TI-RADS classification system designed for papillary thyroid cancer has significant limitations in the differential diagnosis of follicular tumors. Many studies have shown that the area under the curve (AUC) of existing risk stratification systems such as the American College of Radiology (ACR) TI-RADS and Korean TI-RADS for the differential diagnoses of FTA and FTC is only 0.6–0.75, which is significantly lower than their efficiency in the diagnosis of papillary carcinoma (0.85–0.92).15,20 This limitation is mainly attributed to the unique biological behavior of follicular tumors: 1. Malignant features such as capsule invasion and vascular invasion cannot be directly viewed on conventional ultrasound. 2. The morphological characteristics of follicular tumors (such as aspect ratio and margin regularity) significantly overlap between benign and malignant tumors.9,37 Multicenter research data confirmed that even according to the improved French TI-RADS standard, the misclassification rate of follicular tumors remained as high as 30%–40%, resulting in a large number of benign nodules being overbiopsied or operated on unnecessarily.18,38

Pathological mechanisms of overlapping ultrasonographic features
The main reason for the overlap of ultrasonographic features of follicular tumors is associated with their similar histopathological basis. Research shows that FTA and FTC are highly similar in terms of cell arrangement, follicular structure, and other microscopic features. Furthermore, there is no statistical difference in the distribution pattern of collagen fibers that determines their ultrasonographic echogenicity.19,37 It is particularly noteworthy that approximately 25% of FTCs show the phenomenon of “false capsule,” that is, the complete capsule visible on ultrasound exhibits microscopic infiltration on pathological examination. This image–pathology discordance is an important source of misdiagnosis.37,39 In addition, the differences in the angiogenesis patterns between different histological subtypes (such as minimally invasive and extensively invasive FTC) are not fully reflected by conventional color Doppler ultrasound, representing the pathological basis for differential diagnosis.10,19 At the molecular level, the presence of RAS mutation and other co-driving genes contributes to the similarity of ultrasound performance.17,37

Reliability controversy of artificial intelligence (AI)–aided diagnosis
Although the DL system has shown potential in the diagnosis of thyroid nodules, the AI algorithm for follicular tumors continues to face major challenges. Multicenter research data have shown that the accuracy of the DL system based on conventional ultrasound images in distinguishing FTA and FTC is only 68%–72%, which is considerably lower than its performance in the diagnosis of papillary carcinoma (85%–90%).11,19 The main limiting factors include the following: (a) scarcity of training data, because FTC accounts for <15% of all thyroid malignant tumors, resulting in insufficient algorithm learning samples40,41; (b) the interpretation of features is poor, the existing AI model cannot reliably identify key diagnostic features such as capsule discontinuity, and its decision-making process is still in the “black box” state38,42; and (c) clinical transformation barriers. In real world prospective studies, the FDA-approved AI tool showed that the positive predictive value of follicular tumors was only 5%–8% higher than that of radiologists, failing to reach the clinical significance threshold.43,44 Additionally, when the AI system is used in combination with ACR TI-RADS, it may produce a “double misjudgment” effect, that is, the system and classification standard errors are superimposed on each other, considerably reducing the diagnostic accuracy.44,45

Technical limitations and advancement in multimodal fusion

Technical ceiling of conventional ultrasound (resolution/depth limit)
As the preferred method of thyroid nodule screening, conventional ultrasound has obvious technical limitations in the differential diagnosis of follicular tumors. Multicenter research data have shown that the accuracy of the TI-RADS classification system based on conventional ultrasound in differentiating follicular adenoma from FTC is only 60%–70%.
41
The technical bottleneck is mainly reflected in the following: (a) the limitation of spatial resolution makes it difficult to clearly visualize encapsulated microinvasive lesions <2 mm in size
36
; (b) the depth attenuation effect blurs the visualization of the posterior nodule structure and hampers accurate evaluation of the tumor margin
24
; (c) two-dimensional imaging cannot fully reflect the three-dimensional structural characteristics of nodules, resulting in the misclassification of approximately 30% of isoechoic nodules.
10
As a direct effect of these technical defects, up to 43% of benign nodules are unnecessarily recommended for fine-needle aspiration in clinical practice.
37

Incremental value of CEUS and elastography
CEUS technology offers unique advantages in the differential diagnosis of follicular tumors by displaying real-time distribution of micro-vessels. Studies have shown that CEUS can clearly display the unique “fast in and fast out” enhancement mode of FTC, and its diagnostic specificity is 22.5% higher than that of conventional ultrasound.
46
For capsule invasion, the diagnostic sensitivity of CEUS reaches up to 89.3% by displaying an interrupted enhancement ring.
34
Elastography can effectively distinguish the fibrotic areas of FTC by quantifying tissue hardness. Three-dimensional shear wave elastography (3D-SWE) combined with conventional ultrasound can improve the diagnostic accuracy of ACR TI-RADS type 4 nodules to 85.7%.
47
Based on recent research, an algorithm has been developed to convert B-ultrasound images to elastograms so that the application of elastography is no longer limited by special equipment.
48
The multiparameter model shows that the combination of conventional ultrasound, elastography, and CEUS features can increase the differential accuracy for follicular tumors up to 91.2%.
37

Cross validation study of ultrasonomics and genomics
Ultrasonomics provides a new idea for the molecular level identification of follicular tumors by high-throughput extraction of image features. The texture features extracted by the research team from 9393 ultrasound images have shown a significant correlation with the DICER1 gene mutation spectrum.
18
It is particularly noteworthy that ultrasonomics can predict the overexpression of interleukin-6 (IL-6) protein in FTC, with an AUC of 0.81.
35
By analyzing the images of 4317 FTAs and 1593 FTCs, the DL system successfully established a malignant feature recognition model based on random histological sections, with an accuracy rate of 87.6%.
41
Genomics verification found that the deletion of the “halo sign” visible on ultrasound was strongly associated with RAS gene mutation, which provided a molecular basis for diagnosis by correlating imaging features with genetic alterations.
1
At present, the multicenter research is conducting cross-omics matching research of ultrasonographic omics characteristics and TCGA molecular typing.
45

Future research direction and clinical transformation path

Exploration direction for new ultrasonographic biomarkers
Current studies have confirmed that the risk stratification system (RSS) based on ultrasonographic features has obvious limitations in the differential diagnosis of thyroid follicular tumors.
48
Future research should focus on the development of specific ultrasound markers for follicular tumors, including but not limited to the following: (a) micro-texture features extracted based on DL algorithms, such as the fine difference patterns found in the large-scale ultrasound image database (9393 images) analyzed using the overlock network model
49
; (b) combined analysis of three-dimensional ultrasound parameters and molecular markers, especially in terms of characteristics related to vascular invasion (existing data have shown that 62.9% of FTCs have vascular invasion)
32
; (c) research on the correlation between ultrasonomics characteristics and gene mutation spectrum, such as multimodal correlation analysis for the 62.1% mutation detection rate of 66 gene panel detection.
3

Optimization strategy for the multiparameter diagnosis model
Multicenter research data (including 3634 cases from 11 centers) have shown that the existing single-parameter diagnostic criteria have limited efficacy in differentiating follicular adenoma (1748 cases) from cancer (299 cases).
6
The optimization direction should include the following: (a) integration of ultrasound morphological parameters (such as TTW features) and functional parameters (such as elastography values), with existing studies showing a statistically significant difference in the sphericity index between follicular adenoma and cancer
37
; (b) establishment of a decision tree model including clinical risk factors (such as tumor size, average diameter of FTC ≤42 mm)
32
; (c) development of a dynamic risk assessment system to consider the biological behavior differences between different subtypes (such as NIFTP accounting for 4.4%–9.1% of PTC).
7
It is particularly necessary to solve the problem of insufficient applicability of the existing TI-RADS system for follicular tumors.
48

Prospects for the technological integration of ultrasound-guided molecular detection
Regarding the clinical need for preoperative diagnoses, future technology development should focus on the following: (a) the establishment of an ultrasound-guided minimally invasive molecular detection platform, referring to the feasibility of gene panel detection in 1568 cases of multicenter research
3
; (b) combination of intraoperative ultrasound and rapid molecular diagnosis, especially the detection of markers with prognostic significance such as DICER1 mutation
18
; and (c) AI-assisted real-time puncture navigation system that solves the current problem of unnecessary surgery for 44.5% of nonmalignant nodules (including NIFTP and follicular adenoma).
20
The focus needs to break through the limitations of the existing cytological examination for the differential diagnosis of follicular tumors (only 10%–15% of malignant nodules can be diagnosed)
8
and realize the real-time correlation analysis of morphological characteristics and molecular characteristics (Figure 1).

Conclusion
This narrative review underscores the considerable challenges in the preoperative differentiation of FTA from FTC using conventional ultrasonography alone. No single ultrasonographic feature is pathognomonic, and widely used risk stratification systems such as TI-RADS demonstrate suboptimal performance for follicular-patterned lesions. Advanced techniques such as CEUS and elastography provide valuable functional information that can incrementally improve diagnostic confidence, and AI-based models represent a promising future direction. However, current evidence for these adjuncts has mostly been derived from single-center studies that have employed heterogeneous methodologies. Therefore, their translation into routine clinical practice necessitates the development of standardized acquisition protocols, validated diagnostic thresholds, and demonstration of efficacy through large-scale, prospective, multicenter trials. The ultimate goal remains the development of a robust, multiparameter diagnostic framework that integrates imaging as well as clinical and molecular data to minimize diagnostic uncertainty and unnecessary surgery.