Diagnostic performance of the thyroid imaging reporting and data system improved by color-coded acoustic radiation force pulse imaging

Abstract

OBJECTIVE:

To explore the value of color-coded virtual touch tissue imaging (CCV) using acoustic radiation force pulse technology (ARFI) in diagnosing malignant thyroid nodules.

METHODS:

Images including 189 thyroid nodules were collected as training samples and a binary logistic regression analysis was used to calculate regression coefficients for Thyroid Imaging Reporting and Data System (TI-RADS) and CCV. An integrated prediction model (TI-RADS+CCV) was then developed based on the regression coefficients. Another testing dataset involving 40 thyroid nodules was used to validate and compare the diagnostic performance of TI-RADS, CCV, and the integrated predictive models using the receiver operating characteristic (ROC) curves.

RESULTS:

Both TI-RADS and CCV are independent predictors. The diagnostic performance advantage of CCV is insignificant compared to TI-RADS (P = 0.61). However, the diagnostic performance of the integrated prediction model is significantly higher than that of TI-RADS or CCV (all P < 0.05). Applying to the validation image dateset, the integrated predictive model yields an area under the curve (AUC) of 0.880.

CONCLUSIONS:

Developing a new predictive model that integrates the regression coefficients calculated from TI-RADS and CCV enables to achieve the superior performance of thyroid nodule diagnosis to that of using TI-RADS or CCV alone.

Keywords

Elastography acoustic radiation force impulse thyroid imaging reporting and data system thyroid nodules

1 Introduction

Providing patients with appropriate clinical management is inextricably linked to the accurate diagnosis of thyroid nodules [1, 2]. The advantages of real-time, convenience, non-radioactive and non-invasive are why conventional ultrasound (US) is the preferred choice [3, 4]. The US diagnoses thyroid nodules by echogenicity, margins, microcalcifications, etc. The Thyroid Imaging Reporting and Data System (TI-RADS), used by many radiologists, was introduced by the American College of Radiology (ACR) in 2017. Based on TI-RADS, radiologists classified thyroid nodules into five categories (TR1 to TR5) [5 –7]. However, thyroid nodules may be misdiagnosed or missed when benign and malignant nodules present with similar US features. As a result of the extensive malignancy of TR4 and TR5 thyroid nodules, the management options are controversial. Patients may choose to have a fine needle aspiration (FNA) or surgical thyroid nodule removal. Unnecessary FNA or surgery can increase patient anxiety and medical costs [7, 8].

According to previous studies, malignant thyroid nodules often have more stiffness than benign ones. Acoustic radiation force impulse (ARFI) imaging, an elastography technique, can be used to assess the hardness of thyroid nodules. ARFI has been investigated as a diagnostic tool for thyroid nodules, mainly for Virtual Touch Tissue Quantification (VTQ) or Virtual Touch Tissue Imaging and Quantification (VTIQ). VTQ and VTIQ are quantitative methods of ARFI to measure the target tissue’s shear wave velocity (SWV). As the hardness of the tissue increases, the value of SWV also increases. However, it may be that the VTQ or VTIQ often displays X.XX m/s because the shear wave is of poor quality or exceeds the upper limit of the measurement. As a result, VTQ and VTIQ cannot be used widely. It is also possible to assess the stiffness of a nodule at a local level with VTQ or VTIQ but not at the whole nodule level [9 –11].

As a qualitative assessment method for ARFI, virtual touch tissue imaging (VTI) uses acoustic pulses to investigate tissue displacement. VTI has the following advantages: the ROI size can be adjusted according to the size of the nodule; the measurement range is unlimited; it reflects the overall hardness of the nodule; and the hardness of the nodule can be analyzed semi-quantitatively by the scoring method. Using VTI technology, tissue hardness can be displayed in grayscale or color. Studies on VTI have been limited, and most studies use grayscale VTI to diagnose thyroid nodules. Compared to the ability to distinguish different colors, the human eye is significantly weaker in determining the degree of grayscale [12 –16].

Therefore, we aim to investigate the ability of color-coded VTI (CCV) to distinguish benign from malignant thyroid nodules. The added value of CCV for TI-RADS will also be explored simultaneously.

2 Materials and methods

We have received ethical approval for this retrospective study. Due to the anonymity of all patient data, informed consent was not required.

2.1 Patients

From January 2016 to December 2021, we included a group of patients with thyroid nodules in this retrospective study, all of whom underwent US and CCV before surgery or biopsy. The decision to biopsy or operate depends on US findings, patient requests, or risk factors. The results of the CCV test do not affect the decision to perform surgery or biopsy.

The samples for analysis in the training set were 189 thyroid nodules from 159 patients from January 2016 to December 2019. As a validation set, we collected 40 thyroid nodules from January 2021 to December 2021. The flow chart of the study is shown in Fig. 1.

Following are the exclusion criteria: (1) Data from the US and CCV is incomplete; (2) There were incomplete pathological findings in thyroid nodules; (3) Patients who have had a previous thyroidectomy or biopsy of a thyroid nodule; (4) Any cancer patient receiving chemotherapy or radiation therapy.

Fig. 1

Flowchart of the inclusion of patients in the training and validation sets. US = ultrasound. CCV= color-coded virtual touch tissue imaging.

2.2 Instrument

Siemens Acuson S2000 (Siemens Medical Solutions Inc., CA, USA) and equipped with a 4–9 MHz linear transducer, was used as the equipment for US and CCV examinations in this study.

2.3 Image acquisition

All image acquisition was done by an experienced radiologist (an expert in thyroid diagnosis with more than ten years of experience). We first performed a US examination of each thyroid nodule and recorded the nodule’s characteristics, such as size, composition, margins, shape, and calcification. After the US exam was completed, we used the VTI function to check for thyroid nodules. We adjust the size and location of the region of interest (ROI) in VTI mode to include the thyroid nodule and part of the surrounding thyroid tissue. If the nodule’s size exceeded the ROI’s maximum range, we placed the ROI at the junction of the nodule and the surrounding thyroid tissue. In addition to grayscale maps, VTI also displays maps in color. When VTI is displayed in grayscale mode, the tissue in the bright areas is less hard than in the dark places. In the color-coded VTI mode, the order of hardness from high to low is red, orange, green, blue, and purple. Grayscale VTI was not used in this study, only color-coded VTI.

2.4 Image analysis

We randomized all thyroid nodules to be sorted after masking the patient’s name, gender, and age. Two radiologists reviewed the US and CCV images of the thyroid nodules together. Both radiologists had more than five years of experience in thyroid diagnosis. The US and CCV images were not acquired by two radiologists who were unaware of the pathological findings of thyroid nodules. An outcome was reached by discussion when two radiologists disagreed on the definition of image features.

Two radiologists evaluated US features of thyroid nodules according to ACR TI-RADS. In each US feature, a score was assigned, and those with the same score were merged and analyzed. We determined the classification based on the total score obtained for thyroid nodules, as follows: benign (0 points, TR1); no suspicion (2 points, TR2); mildly suspicious (3 points, TR3); moderately suspicious (4–6 points, TR4); and highly suspicious (7 points or more, TR5) [17].

In CCV, we assessed the hardness of both the nodule and the surrounding tissue. In this study, CCV scores were performed for each thyroid nodule as follows: If no red or orange is seen in the nodule, indicating low hardness (score 1); A nodule with an orange primary color and a green secondary color is rated as medium hard (score 2); Score 3 were assigned to nodules whose predominant colors were red and orange, followed by other colors, indicating high hardness; When nodules and surrounding tissues have a dominant color of red or orange, the nodules and surrounding tissues are considered to be hard (score 4) [15].

2.5 Interobserver agreement analysis

We randomly selected 30 cases of thyroid nodules for interobserver agreement analysis. Two radiologists with different experiences performed TI-RADS classification and CCV score on 30 thyroid nodules, for which they were unaware of the pathological findings. Both radiologists had six and three years of experience, respectively.

2.6 Statistical analysis

In this study, we used MedCalc (version 18.0) and SPSS software (version 26.0) for statistical analysis. Count data were analyzed using the chi-square test or Fisher’s exact test. We analyzed continuous variables using a t-test or Mann-Whitney U-test. Spearman’s rank correlation was used to assess the correlation between TI-RADS and CCV. We developed the logistic regression equation based on a binary logistic regression with TI-RADS and CCV as independent variables and pathological outcomes as dependent variables.

$Logit (P) = X 0 + X 1 \times CCV + X 2 \times TI-RADS$

The equation includes are two regression coefficients X1 and X2, and one constant X0. The Hosmer and Lemeshow test was used to test the model’s goodness of fit. The integrated prediction model (TI-RADS+CCV) was defined as TI-RADS+X1/X2×CCV [14 , 18]. An evaluation and comparison of TI-RADS, CCV, and integrated prediction model diagnostic performance were conducted using receiver operating characteristic curves (ROC). The validation set data were used to validate the diagnostic performance of the integrated prediction model [19, 20]. We included 40 thyroid nodules in the validation set, which met the same criteria as the training set. We assessed interobserver agreement using the Kappa statistic (poor agreement, K-value less than 0.40; moderate agreement, K-value between 0.40 and 0.60; substantial agreement, K-value between 0.60 and 0.80; and almost perfect agreement, K-value greater than 0.80). Comparisons between independent proportions were made using chi-square tests. Differences were statistically significant if the P-value was less than 0.05.

3 Results

3.1 Participant characteristics

In this study, 199 patients with 229 thyroid nodules were included. Table 1 summarizes the essential features, TI-RADS classification, and CCV scores of patients with thyroid nodules. The training set (n = 189) and validation set (n = 40) did not differ significantly in essential features, TI-RADS classification, or CCV scores (P > 0.05).

Table 1
Basic characteristics, TI-RADS, and CCV in the training and validation sets

Parameters Training set (n = 189) Validation set (n = 40) P

Gender 0.577^a

Female 164 (86.8) 36 (90.0)

Male 25 (13.2) 4 (10.0)

Age (y) 46.23±14.56 45.28±14.29 0.707^c

Size (mm) 25.98±15.07 28.79±12.92 0.273^c

Histological diagnosis 0.805^a

Benign 143 (75.7) 31 (77.5)

Malignant 46 (24.3) 9 (22.5)

TI-RADS 0.347^a

TR2 43 (22.8) 13 (32.5)

TR3 42 (22.2) 11 (27.5)

TR4 58 (30.7) 10 (25.0)

TR5 46 (24.3) 6 (15.0)

CCV 0.487^b

Score 1 26 (13.8) 6 (15.0)

Score 2 93 (49.2) 17 (42.5)

Score 3 46 (24.3) 14 (35.0)

Score 4 24 (12.7) 3 (7.5)

Parameters	Training set (n = 189)	Validation set (n = 40)	P
Gender			0.577^a
Female	164 (86.8)	36 (90.0)
Male	25 (13.2)	4 (10.0)
Age (y)	46.23±14.56	45.28±14.29	0.707^c
Size (mm)	25.98±15.07	28.79±12.92	0.273^c
Histological diagnosis			0.805^a
Benign	143 (75.7)	31 (77.5)
Malignant	46 (24.3)	9 (22.5)
TI-RADS			0.347^a
TR2	43 (22.8)	13 (32.5)
TR3	42 (22.2)	11 (27.5)
TR4	58 (30.7)	10 (25.0)
TR5	46 (24.3)	6 (15.0)
CCV			0.487^b
Score 1	26 (13.8)	6 (15.0)
Score 2	93 (49.2)	17 (42.5)
Score 3	46 (24.3)	14 (35.0)
Score 4	24 (12.7)	3 (7.5)

a = Determined with the χ² test. b = Determined with Fisher’s exact test. c = Determined with the t-test. Notes. Unless otherwise noted, data points are the number of lesions, and numbers in parentheses are percentages. TI-RADS = thyroid imaging report and data system. CCV = color-coded virtual touch tissue imaging.

3.2 Pathological diagnosis of all thyroid nodules

The training set included 46 malignant and 143 benign nodes, all confirmed by biopsy or surgery. The 46 malignant nodules included 43 papillary carcinomas and three follicular adenocarcinomas. And among 143 benign nodules, 129 were nodular goiter, 13 were follicular adenoma, and 1 was subacute thyroiditis.

The validation set included nine malignant and 31 benign nodes, all confirmed by biopsy or surgery. The nine malignant nodules had eight papillary carcinomas and one follicular adenocarcinoma. And among the 31 benign nodules, 26 were nodular goiter, and 5 were follicular adenomas.

3.3 The univariate analyses for the training set

Table 2 shows the US characteristics and CCV scores of thyroid nodules. Significant differences in US features between benign and malignant nodules included echogenicity, composition, margins, shape, and microcalcifications (all P < 0.001). In this study, malignant thyroid nodules were predominantly TR5 (P < 0.001). Benign nodules had a significantly lower CCV score than malignant nodules (P < 0.001) (Figs. 2 –5).

Table 2
Sonographic features, TI-RADS, and CCV of benign and malignant thyroid nodules in the training dataset

Parameters Histological diagnosis P

Benign (n = 143) Malignant (n = 46)

Composition <0.001^a

Mixed cystic and solid 61 (42.7) 4 (8.7)

Solid or almost solid 82 (57.3) 42 (91.3)

Echogenicity <0.001^b

Iso/Mix-echoic 71 (49.7) 4 (8.7)

Hypo-echoic 69 (48.3) 38 (82.6)

Very hypo-echoic 3 (2.1) 4 (8.7)

Shape (A/T) <0.001^a

<1 135 (94.4) 26 (56.5)

>1 8 (5.6) 20 (43.5)

Margin <0.001^a

Smooth or ill-defined 132 (92.3) 10 (21.7)

Lobulated or irregular 11 (7.7) 36 (78.3)

Calcification <0.001^b

None/Large comet-tail artifacts 128 (89.5) 11 (23.9)

Macrocalcification 8 (5.6) 6 (13.1)

Microcalcification 7 (4.9) 29 (63.0)

TI-RADS <0.001^a

2 (TR 2) 43 (30.1) 0 (0.0)

3 (TR 3) 39 (27.3) 3 (6.5)

4 (TR 4) 50 (35.0) 8 (17.4)

5 (TR 5) 11 (7.7) 35 (76.1)

CCV <0.001^a

1 (Score 1) 26 (18.2) 0 (0.0)

2 (Score 2) 88 (61.5) 5 (10.9)

3 (Score 3) 29 (20.3) 17 (37.0)

4 (Score 4) 0 (0.0) 24 (52.2)

TI-RADS+CCV 6.6375±1.6053 10.4962±1.4118 <0.001^b

Benign (≤7.3991) 115 (80.4) 1 (2.2)

Malignant (>7.3991) 28 (19.6) 45 (97.8)

Parameters	Histological diagnosis	P
Composition			<0.001^a
Mixed cystic and solid	61 (42.7)	4 (8.7)
Solid or almost solid	82 (57.3)	42 (91.3)
Echogenicity			<0.001^b
Iso/Mix-echoic	71 (49.7)	4 (8.7)
Hypo-echoic	69 (48.3)	38 (82.6)
Very hypo-echoic	3 (2.1)	4 (8.7)
Shape (A/T)			<0.001^a
<1	135 (94.4)	26 (56.5)
>1	8 (5.6)	20 (43.5)
Margin			<0.001^a
Smooth or ill-defined	132 (92.3)	10 (21.7)
Lobulated or irregular	11 (7.7)	36 (78.3)
Calcification			<0.001^b
None/Large comet-tail artifacts	128 (89.5)	11 (23.9)
Macrocalcification	8 (5.6)	6 (13.1)
Microcalcification	7 (4.9)	29 (63.0)
TI-RADS			<0.001^a
2 (TR 2)	43 (30.1)	0 (0.0)
3 (TR 3)	39 (27.3)	3 (6.5)
4 (TR 4)	50 (35.0)	8 (17.4)
5 (TR 5)	11 (7.7)	35 (76.1)
CCV			<0.001^a
1 (Score 1)	26 (18.2)	0 (0.0)
2 (Score 2)	88 (61.5)	5 (10.9)
3 (Score 3)	29 (20.3)	17 (37.0)
4 (Score 4)	0 (0.0)	24 (52.2)
TI-RADS+CCV	6.6375±1.6053	10.4962±1.4118	<0.001^b
Benign (≤7.3991)	115 (80.4)	1 (2.2)
Malignant (>7.3991)	28 (19.6)	45 (97.8)

a = Determined with the χ² test. b = Determined with Fisher’s exact test. Notes. Unless otherwise noted, data points are the number of lesions, and numbers in parentheses are percentages. A/T = anteroposterior/transverse diameter. TI-RADS = thyroid imaging report and data system. CCV = color-coded virtual touch tissue imaging.

Fig. 2

Comparison of conventional US (left) and CCV (right) images of a follicular adenoma with a CCV score of 1 and a TI-RADS classification of TR3. US = ultrasound. CCV= color-coded virtual touch tissue imaging. TI-RADS = thyroid imaging reporting and data system.

Fig. 3

Comparison of conventional US (left) and CCV (right) images of a case of nodular goiter with a CCV score of 2 and a TI-RADS classification of TR3. US = ultrasound. CCV= color-coded virtual touch tissue imaging. TI-RADS = thyroid imaging reporting and data system.

Fig. 4

Comparison of conventional US (left) and CCV (right) images of papillary thyroid carcinoma with a CCV score of 3 and a TI-RADS classification of TR5. US = ultrasound. CCV= color-coded virtual touch tissue imaging. TI-RADS = thyroid imaging reporting and data system.

Fig. 5

Comparison of conventional US (left) and CCV (right) images of papillary thyroid carcinoma with a CCV score of 4 and a TI-RADS classification of TR5. US = ultrasound. CCV= color-coded virtual touch tissue imaging. TI-RADS = thyroid imaging reporting and data system.

3.4 Correlation between TI-RADS and CCV

In the evaluation of thyroid nodules, a strong correlation was found between TI-RADS and CCV (R = 0.710, P < 0.001).

3.5 Multivariate analysis of TI-RADS and CCV in the training set

Both TI-RADS (odds ratio [OR]: 3.736; 95% confidence interval [CI]: 1.775, 7.863) and CCV (OR: 9.397; 95% CI: 3.620, 24.396) were independent predictors of malignant thyroid nodules(all P < 0.05) (Table 3). The regression equation was written as follows:

Table 3
Multivariate logistic regression analysis based on TI-RADS and CCV in the training dataset

Parameters B SE Wald Exp(B) OR 95% CI P

Lower limit Upper limit

TI-RADS 1.318 0.380 12.050 3.736 1.775 7.863 0.001

CCV 2.240 0.487 21.185 9.397 3.620 24.396 <0.001

Constant –12.524 1.899 43.508

Parameters	B	SE	Wald	Exp(B)	OR 95% CI	P
TI-RADS	1.318	0.380	12.050	3.736	1.775	7.863	0.001
CCV	2.240	0.487	21.185	9.397	3.620	24.396	<0.001
Constant	–12.524	1.899	43.508

TI-RADS = thyroid imaging report and data system. CCV = color-coded virtual touch tissue imaging. CI = confidence intervals. B = regression coefficient. SE = standard error. Wald = chi-square value. The Hosmer & Lemeshow test was 0.893.

Table 4

Diagnostic performance of TI-RADS, CCV, or both for thyroid nodules in the training dataset

Data^*	Cut-off	AUC (95% CI)	SEN(%)	SPE(%)
TI-RADS	>TR4	0.890^c (0.841–0.940)	76.09^c [35/46]	92.31^bc [132/143]
CCV	>2	0.907^c (0.859–0.955)	89.13 [41/46]	79.72^a [114/143]
TI-RADS+CCV	>7.3991	0.954^ab (0.928–0.980)	97.83^a [45/46]	80.42^a [115/143]

Note. Unless otherwise indicated, the data point is a percentage, and the number in parentheses is 95% CI. TI-RADS = thyroid imaging report and data system. CCV = color-coded virtual touch tissue imaging. CI = confidence intervals. SEN = sensitivity. SPE = specificity. * = data are percentages and raw data in brackets. a = Compared with TI-RADS, P < 0.05. b = Compared with CCV, P < 0.05. c = Compared with TI-RADS+CCV, P < 0.05.

$Logit (P) = - 12.524 + 2.240 \times CCV + 1.318 \times TI - RADS$

Additionally, Hosmer and Lemeshow test had a value of 0.893.

3.6 Diagnostic performance of TI-RADS, CCV, and the integrated predictive model in the training set

In the present study, the area under the curve (AUC) was higher for CCV than for TI-RADS, but the difference was insignificant (P = 0.6104). The integrated prediction model had the largest AUC, whether compared with TI-RADS or CCV (all P < 0.05) (Fig. 6).

Fig. 6

ROC analysis of TI-RADS, CCV, and TI-RADS+CCV in the training set. ROC = receiver operating characteristic curves. TI-RADS = thyroid imaging reporting and data system. CCV = color-coded virtual touch tissue imaging.

3.7 Validation of the integrated predictive model

Table 5 shows the variables used in the validation set for the integrated prediction model. The integrated prediction model performed equally well for diagnosis in the validation set, with AUC, sensitivity, and specificity of 0.880, 88.89%, and 77.42%, respectively (Table 6) (Fig. 7).

Table 5
TI-RADS, CCV, or both for thyroid nodules in the validation dataset

Parameters Histological diagnosis P

Benign (n = 31) Malignant (n = 9)

TI-RADS 0.007^a

2 (TR 2) 13 (41.9) 0 (0.0)

3 (TR 3) 9 (29.0) 2 (22.2)

4 (TR 4) 7 (22.6) 3 (33.3)

5 (TR 5) 2 (6.5) 4 (44.5)

CCV <0.001^a

1 (Score 1) 6 (19.4) 0 (0.0)

2 (Score 2) 20 (64.5) 1 (11.1)

3 (Score 3) 5 (16.1) 5 (55.6)

4 (Score 4) 0 (0.0) 3 (33.3)

TI-RADS+CCV 10.1634±2.7047 16.0581±3.1553 <0.001^b

Benign (≤7.3991) 24 (77.4) 1 (11.1)

Malignant (>7.3991) 7 (22.6) 8 (88.9)

Parameters	Histological diagnosis	P
TI-RADS			0.007^a
2 (TR 2)	13 (41.9)	0 (0.0)
3 (TR 3)	9 (29.0)	2 (22.2)
4 (TR 4)	7 (22.6)	3 (33.3)
5 (TR 5)	2 (6.5)	4 (44.5)
CCV			<0.001^a
1 (Score 1)	6 (19.4)	0 (0.0)
2 (Score 2)	20 (64.5)	1 (11.1)
3 (Score 3)	5 (16.1)	5 (55.6)
4 (Score 4)	0 (0.0)	3 (33.3)
TI-RADS+CCV	10.1634±2.7047	16.0581±3.1553	<0.001^b
Benign (≤7.3991)	24 (77.4)	1 (11.1)
Malignant (>7.3991)	7 (22.6)	8 (88.9)

a = Determined with Fisher’s exact test. b = Determined with the t-test. Notes. Unless otherwise noted, data points are the number of lesions, and numbers in parentheses are percentages. TI-RADS = thyroid imaging report and data system. CCV = color-coded virtual touch tissue imaging.

Table 6

Diagnostic performance of TI-RADS+CCV in the validation dataset

Data^*	AUC (95% CI)	SEN(%)	SPE(%)
TI-RADS+ CCV	0.880 (0.765–0.995)	88.89 [8/9]	77.42 [24/31]

Note. Unless otherwise indicated, the data point is a percentage, and the number in parentheses is 95% CI. TI-RADS = thyroid imaging report and data system; CCV = color-coded virtual touch tissue imaging; CI = confidence intervals; SEN = sensitivity; SPE = specificity. * = data are percentages and raw data in brackets.

Fig. 7

ROC analysis of TI-RADS+CCV in the validation set. ROC = receiver operating characteristic curves. TI-RADS = thyroid imaging reporting and data system. CCV = color-coded virtual touch tissue imaging.

3.8 Interobserver agreement analysis

Our results show that radiologists with different experiences can obtain substantial agreement using the TI-RADS (K-value: 0.658 and 95% CI: 0.442, 0.874) and CCV (K-value: 0.633 and 95% CI: 0.396, 0.869).

4 Discussion

In the training set, malignant nodules were mainly scored as 3 and 4, while benign nodules were mainly scored as 1 and 2. When the hardness of thyroid nodules and surrounding tissues increases, the nodule may be malignant. Because the invasion of cancer cells may increase the hardness of the surrounding tissues. The CCV score effectively diagnosed malignant thyroid nodules with a sensitivity, specificity, and AUC of 89.13%, 79.72%, and 0.907, respectively. Our results are consistent with those of previous elastography studies [21, 22].

Conventional US remains the primary method for diagnosing thyroid nodules despite the presence of elastography. In the present study, the area under the curve (AUC) was higher for CCV than for TI-RADS, but the difference was insignificant (P = 0.6104), similar to previous studies [23, 24].

There were limitations when the CCV was used alone. The results of CCV may vary based on the pulsations of the large blood vessels in the neck. The CCV does not provide information about the thyroid nodule or surrounding tissue other than its hardness. Benign nodules and malignant nodules can have the same level of hardness.

The TI-RADS classification is a semi-quantitative comprehensive evaluation method that classifies thyroid nodules based on US characteristics. Malignant thyroid nodules that are small in size and have an aspect ratio of less than 1, or benign nodules with internal calcification, can be misdiagnosed by the TI-RADS classification alone.

The single application of CCV or TI-RADS classification produces a high rate of misdiagnosis, whereas the combined application allows for complementarity and improved diagnostic performance. Based on the results of previous studies, the combination of elastography and conventional US could improve the diagnostic accuracy of thyroid nodules [25 –27]. Our results showed that the combined prediction model had the largest AUC (all P < 0.05) compared to either TI-RADS or CCV.

In this study, we generated a prediction model (TI-RADS+CCV) using the training set data and validated them using an external validation set. The validation set does not contain duplicate data, and the training set and validation set do not overlap. In the validation set, TI-RADS+CCV showed good diagnostic power with a sensitivity of 88.89%, specificity of 77.42%, and AUC of 0.880.

There were the following limitations in this study. Our study was a retrospective study and might have some selectivity bias. We only included patients with thyroid nodules who underwent surgical resection or biopsy for the study and, therefore, might have resulted in a higher malignancy rate that was not representative of the overall population incidence. The proportion of malignant nodule cases was small, and the pathological types were unevenly distributed; in the follow-up study, we will expand the sample size and increase the pathological types. The same radiologist did all acquisitions of conventional US images and CCV images in our study. There was no comparison of differences in image acquisition by different radiologists. Similarly, the diagnostic performance of other radiologists who use CCV was not analyzed, and future studies should address and improve these issues.

5 Conclusion

Based on the results of this study, we thus recommend CCV as an adjunct to TI-RADS to improve diagnosis because CCV provides additional information on hardness.

Footnotes

Acknowledgments

The authors gratefully acknowledge Wenxia Lin, Xiaohuan Zhu, and Xiao-E Huang from the Department of Ultrasound, The First Affiliated Hospital of Shantou University Medical College, Shantou, Guangdong Province, P.R. China.

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

This retrospective study was applied for review and approved by the Institutional Review Board (The first affiliated Hospital of Medical College of Shantou University-Ethical review-No. B-2021-020).

Informed consent

The institutional review board approved waivers of informed consent.

Funding

This study was supported by the Shantou Bureau of Science and Technology (NO. 210413116491209).

References

Armstrong

M.J.

, Yang

, Yip

et al Pax8/pparγ rearrangement in thyroid nodules predicts follicular-pattern carcinomas, in particular the encapsulated follicular variant of papillary carcinoma, Thyroid 24(9) (2014), 1369–1374.

Tumino

, Grani

, Di Stefano

et al Nodular thyroid disease in the era of precision medicine, Front Endocrinol (Lausanne) 10 (2019), 907.

, Duan

, Zhang

et al. Efficient deep learning architecture for detection and recognition of thyroid nodules, Comput Intell Neurosci 2020 (2020), 1242781.

Chen

H.Y.

, Liu

W.Y.

, Zhu

et al. Diagnostic value of contrast-enhanced ultrasound in papillary thyroid microcarcinoma, Exp Ther Med 11(5) (2016), 1555–1562.

Pei

, Zhang

, Cong

et al. Ultrasound real-time tissue elastography improves the diagnostic performance of the acr thyroid imaging reporting and data system in differentiating malignant from benign thyroid nodules: A summary of thyroid nodules, Int J Endocrinol 2020 (2020), 1749351.

Barbosa

T.L.M.

, Junior

C.O.M.

, Graf

et al. Acr ti-rads and ata us scores are helpful for the management of thyroid nodules with indeterminate cytology, BMC Endocr Disord 19(1) (2019), 112.

Wang

, He

, Ma

et al. Multimode ultrasonic technique is recommended for the differential diagnosis of thyroid cancer, PeerJ 8 (2020), e9112.

Deng

X.H.

, Tang

L.N.

, Liu

S.Q.

et al. A proposal to stratify the intermediate-risk thyroid nodules according to the aace/ace/ame guidelines with ultrasound features, Sci Rep 7(1) (2017), 17901.

Zhang

, Zhang

, Zheng

et al. Acoustic radiation force impulse imaging for the differentiation of benign and malignant lymph nodes: A systematic review and meta-analysis, PLoS One 11(11) (2016), e0166716.

10.

Liang

, Wu

, Shang

et al. Vtiq evaluates antitumor effects of net-1 sirna by utmd in hcc xenograft models, Oncol Lett 16(3) (2018), 2893–2902.

11.

Golatta

, Schweitzer-Martin

, Harcos

et al. Evaluation of virtual touch tissue imaging quantification, a new shear wave velocity imaging method, for breast lesion assessment by ultrasound, Biomed Res Int 2014 (2014), 960262.

12.

Bai

, Zhang

H.P.

, Xing

J.F.

et al. Acoustic radiation force impulse technology in the differential diagnosis of solid breast masses with different sizes: Which features are most efficient? Biomed Res Int 2015 (2015), 410560.

13.

Lian

K.M.

, Lin

The value of color-map virtual touch tissue imaging (cmv) in the differential diagnosis of benign and malignant breast lesions, Clin Hemorheol Microcirc 78(1) (2021), 49–56.

14.

Lian

K.M.

, Lin

Color-map virtual touch tissue imaging (cmv) combined with bi-rads for the diagnosis of breast lesions, J Xray Sci Technol 30(3) (2022), 447–457.

15.

Lian

K.M.

, Lin

Role of color-coded virtual touch tissue imaging in suspected thyroid nodules, Technol Health Care 30(3) (2022), 673–682.

16.

Zhou

, Yang

, Zhan

et al. Breast lesions evaluated by color-coded acoustic radiation force impulse (arfi) imaging, Ultrasound Med Biol 42(7) (2016), 1464–1472.

17.

Tessler

F.N.

, Middleton

W.D.

, Grant

E.G.

et al. ACR thyroid imaging, reporting and data system (ti-rads): White paper of the ACR TI-RADS committee, J Am Coll Radiol 14(5) (2017), 587–595.

18.

Pei

, Zhang

, Cong

et al. Ultrasound real-time tissue elastography improves the diagnostic performance of the ACR thyroid imaging reporting and data system in differentiating malignant from benign thyroid nodules: A summary of thyroid nodules, Int J Endocrinol 2020 (2020), 1749351.

19.

Cai

, Wang

, Zhang

et al. Superb microvascular imaging technology can improve the diagnostic efficiency of the BI-RADS system, Front Oncol 11 (2021), 634752.

20.

Wang

, Dong

, Nie

et al. Contrast-enhanced ultrasound in the differential diagnosis and risk stratification of ACR TI-RADS category 4 and 5 thyroid nodules with non-hypovascular, Front Oncol 11 (2021), 662273.

21.

Liu

B.J.

, Lu

, Xu

H.X.

et al. The diagnosis value of acoustic radiation force impulse (arfi) elastography for thyroid malignancy without highly suspicious features on conventional ultrasound, Int J Clin Exp Med 8(9) (2015), 15362–15372.

22.

Zhang

Y.F.

, Liu

, Xu

H.X.

et al. Acoustic radiation force impulse imaging: A new tool for the diagnosis of papillary thyroid microcarcinoma, Biomed Res Int 2014 (2014), 416969.

23.

Borlea

, Borcan

, Sporea

et al. TI-RADS diagnostic performance: Which algorithm is superior and how elastography and 4D vascularity improve the malignancy risk assessment, Diagnostics (Basel) 10(4) (2020), 180.

24.

, Qu

, Zang

et al. Preliminary study of confounding factors of elastography and the application of fine-needle aspiration in thyroid nodules with indeterminate elastography, Sci Rep 7(1) (2017), 18005.

25.

Chen

B.D.

, Xu

H.X.

, Zhang

Y.F.

et al. The diagnostic performances of conventional strain elastography (se), acoustic radiation force impulse (arfi) imaging and point shear-wave speed (psws) measurement for non-calcified thyroid nodules, Clin Hemorheol Microcirc 65(3) (2017), 259–273.

26.

Magri

, Chytiris

, Chiovato

The role of elastography in thyroid ultrasonography, Curr Opin Endocrinol Diabetes Obes 23(5) (2016), 416–422.

27.

Bardet

, Ciappuccini

, Pellot-Barakat

et al. Shear wave elastography in thyroid nodules with indeterminate cytology: Results of a prospective bicentric study, Thyroid 27(11) (2017), 1441–1449.

Parameters	Histological diagnosis		P
	Benign (n = 143)	Malignant (n = 46)
Composition			<0.001^a
Mixed cystic and solid	61 (42.7)	4 (8.7)
Solid or almost solid	82 (57.3)	42 (91.3)
Echogenicity			<0.001^b
Iso/Mix-echoic	71 (49.7)	4 (8.7)
Hypo-echoic	69 (48.3)	38 (82.6)
Very hypo-echoic	3 (2.1)	4 (8.7)
Shape (A/T)			<0.001^a
<1	135 (94.4)	26 (56.5)
>1	8 (5.6)	20 (43.5)
Margin			<0.001^a
Smooth or ill-defined	132 (92.3)	10 (21.7)
Lobulated or irregular	11 (7.7)	36 (78.3)
Calcification			<0.001^b
None/Large comet-tail artifacts	128 (89.5)	11 (23.9)
Macrocalcification	8 (5.6)	6 (13.1)
Microcalcification	7 (4.9)	29 (63.0)
TI-RADS			<0.001^a
2 (TR 2)	43 (30.1)	0 (0.0)
3 (TR 3)	39 (27.3)	3 (6.5)
4 (TR 4)	50 (35.0)	8 (17.4)
5 (TR 5)	11 (7.7)	35 (76.1)
CCV			<0.001^a
1 (Score 1)	26 (18.2)	0 (0.0)
2 (Score 2)	88 (61.5)	5 (10.9)
3 (Score 3)	29 (20.3)	17 (37.0)
4 (Score 4)	0 (0.0)	24 (52.2)
TI-RADS+CCV	6.6375±1.6053	10.4962±1.4118	<0.001^b
Benign (≤7.3991)	115 (80.4)	1 (2.2)
Malignant (>7.3991)	28 (19.6)	45 (97.8)

Parameters	Histological diagnosis		P
	Benign (n = 31)	Malignant (n = 9)
TI-RADS			0.007^a
2 (TR 2)	13 (41.9)	0 (0.0)
3 (TR 3)	9 (29.0)	2 (22.2)
4 (TR 4)	7 (22.6)	3 (33.3)
5 (TR 5)	2 (6.5)	4 (44.5)
CCV			<0.001^a
1 (Score 1)	6 (19.4)	0 (0.0)
2 (Score 2)	20 (64.5)	1 (11.1)
3 (Score 3)	5 (16.1)	5 (55.6)
4 (Score 4)	0 (0.0)	3 (33.3)
TI-RADS+CCV	10.1634±2.7047	16.0581±3.1553	<0.001^b
Benign (≤7.3991)	24 (77.4)	1 (11.1)
Malignant (>7.3991)	7 (22.6)	8 (88.9)

Diagnostic performance of the thyroid imaging reporting and data system improved by color-coded acoustic radiation force pulse imaging

Abstract

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1 Introduction

2 Materials and methods

2.1 Patients

2.3 Image acquisition

2.4 Image analysis

2.5 Interobserver agreement analysis

2.6 Statistical analysis

3 Results

3.1 Participant characteristics

3.3 The univariate analyses for the training set

3.5 Multivariate analysis of TI-RADS and CCV in the training set

Table 3 Multivariate logistic regression analysis based on TI-RADS and CCV in the training dataset Parameters B SE Wald Exp(B) OR 95% CI P Lower limit Upper limit TI-RADS 1.318 0.380 12.050 3.736 1.775 7.863 0.001 CCV 2.240 0.487 21.185 9.397 3.620 24.396 <0.001 Constant –12.524 1.899 43.508

4 Discussion

5 Conclusion

Footnotes

Acknowledgments

Conflict of interest

Ethical approval

Informed consent

Funding

References

Table 3
Multivariate logistic regression analysis based on TI-RADS and CCV in the training dataset

Parameters B SE Wald Exp(B) OR 95% CI P

Lower limit Upper limit

TI-RADS 1.318 0.380 12.050 3.736 1.775 7.863 0.001

CCV 2.240 0.487 21.185 9.397 3.620 24.396 <0.001

Constant –12.524 1.899 43.508