The clinical value of Arrival-time Parametric Imaging using contrast-enhanced ultrasonography in differentiating benign and malignant breast lesions

Abstract

OBJECTIVES:

To evaluate the clinical value of Arrival-time Parametric Imaging (At-PI) in the differentiation of benign and malignant breast lesions.

METHODS:

For this ethics committee-approved retrospective study, a total of 184 breast lesions in 176 women were included and gray-scale ultrasound, contrast-enhanced ultrasound (CEUS) and At-PI were performed. In CEUS and At-PI, perfusion patterns, perfusion uniformity and color spatial distribution for lesions were analyzed qualitatively and the maximal diameter ratio of the lesion in accumulated parametric images and that in gray-scale images (MDR_AI/GI) and area ratio of the lesion in accumuated parametric images and that in gray-scale images (AR_AI/GI) were calculated quantitatively. Kappa and Intraclass Correlation Coefficient were used to evaluate the interobserver reproducibility for CEUS and At-PI and the intraobserver reproducibility for At-PI, respectively. The area under receiver operating characteristic (AUC), sensitivity, specificity, accuracy and positive and negative likelihood ratios (PPV, NPV) were calculated for MDR_AI/GI and AR_AI/GI.

RESULTS:

Good interobserver and intraobserver reproducibility for At-PI were identified. In At-PI, there were statistically significant differences in perfusion patterns, color spatial distribution, MDR_AI/GI and AR_AI/GI between benign and malignant breast lesions (P < 0.05). The AUCs of MDR_AI/GI and AR_AI/GI were 0.895 and 0.954, respectively, with no significant difference between them (Z = 1.84, P > 0.05). By using the thresholds of 1.125 for MDR_AI/GI and 1.21 for AR_AI/GI, the sensitivity, specificity, accuracy, PPV and NPV of At-PI were 84.48%, 88.24%, 85.57%, 92.45% and 76.92%, respectively, for MDR_AI/GI and 93.10%, 91.18%, 92.39%, 94.74% and 88.57%, respectively, for AR_AI/GI.

CONCLUSIONS:

At-PI is helpful to distinguish benign from malignant breast lesions. And MDR_AI/GI and AR_AI/GI are useful and efficient features for differential diagnosis.

Keywords

Contrast-enhanced ultrasound Arrival-time Parametric Imaging benign and malignant breast lesions

1 Introduction

Breast cancer is a common malignancy and one of the greatest threasts to women’s health. At present, it has been demonstrated that contrast-enhanced ultrasound (CEUS) is helpful to differentiate benign and malignant breast lesions. And range increase of breast lesions in contrast-enhanced ultrasound images on peak time (CI) is one of the malignant features [1 –4]. Yuxin Jiang et al. [5] reported that the size of malignant lesions increased 3 to 16 mm (mean: 5 mm) in CI, compared to gray-scale images (GI). However, to date there has been no uniform diagnostic criteria to distinguish benign from malignant breast lesions on range increase by CEUS.

Arrival-time Parametric Imaging (At-PI), as a time-dependent imaging technology, displays graduated colors from red to purple on dynamic contrast-enhanced ultrasound images. It is a useful method to reconstrust vessel structure by adding temporal information of the inflow. Currently, some articles have reported that At-PI was helpful to diagnose liver diseases [6, 7]. However, until now few articles have focused on the application of At-PI in breast diseases. Aya Noro et al. [8] assessed the visibility of accumulatrd parametric images (AI) and demonstrated that the maximal diameters of tumors determined in AI correlated well with the pathology.

As far as we know, no study has evaluated the clinical value of At-PI in the differential diagnosis of breast lesions. In this study, the maximal diameter ratio of the lasion in AI and that in GI (MDR_AI/GI) and the area ratio of the lesion in AI and that in GI (AR_AI/GI) are first considered as features to evaluate the diagnostic performance of At-PI in the diagnosis of breast lesions.

2 Materials and methods

2.1 Patients

This retrospective research was approved by the Ethics Committee of our hospital and written informed consents of all patients were obtained before CEUS and surgery or percutaneous biopsy. From January 2017 to November 2019, a total of 202 breast lesions (BI-RADS categories 3–5) in 194 women were included in our study. All patients were performed CEUS and pathologic results were obtained. However, 18 patients with 18 lesions were excluded. 9 of the 18 lesions were excluded for the unstable dynamic contrast-enhanced ultrasound images or the lack of CEUS data. 7 were excluded for bulky calcification with ultasound attenuation. And 2 were excluded because lesions were too big to be observed in one plane. Finally, 184 breast lesions in 176 patients (age range, 21–75 years; mean±SD, 48±14 years) were included and assessed in our study. The pathologic results revealed 116 benign lesions (the mean maximal diameter±SD in GI, 13.72±7.61 mm; the area±SD in GI, 1.05±1.53 cm²) and 68 malignant lesions (the mean maximal diameter±SD in GI, 22.76±13.26 mm; the area±SD in GI, 3.20±4.93 cm²).

2.2 Contrast-enhanced ultrasound

All the examinations were performed before surgery or percutaneous biopsy. LOGIQ E9 (GE Healthcare, Milwaukee, WI) with 9L probe (5–9 MHz) was used for gray-scale ultrasound (gray-scale US) and CEUS examinations with a mechanical index of 0.9–1.2, dynamic range 48–60 dB and 12–22 dB gain set was used. The contrast agent was SonoVue (Bracco, Milan, Italy), a sulphur hexafluride microbubble contrast agent. And a suspension consisted of sulphur hexafluride microbubbles powder and 5 mL saline solution for preparation. All the patients were in the supine position and fully exposed their breasts. CEUS was performed by the same physician with more than 5 years’ experience in breast CEUS. Before CEUS, we selected the ideal plane with the maximal diameter based on gray-scale US. Then dual-image contrast mode was switched to during the whole examination. After a bolus injection of 4.5 ml contrast agent at the elbow vein followed by 5 ml saline solution, a two minutes dynamic image was recorded and saved as raw data on a hard disk for further analysis.

2.3 Parametric imaging

The built-in At-PI software in LOGIQ E9 was used with the same settings of machine parameters as CEUS. In At-PI, different colors were used to represent different arrival times of sulphur hexafluride microbubbles from red to orange, yellow, green, blue and purple. And the time when sulphur hexafluride microbubbles reached lesions was set as the starting point. In the graduated color scale, the time interval between each two colors is 2 seconds. It means that red represents the earlist arrival time and purple represents the latest arrival time in lesions. Subsequently, dynamic images were saved. To avoid interobserver variability, At-PI and subsequent qualitative and quantitative analysis were performed by another expert (he was not the physician who performed gray-scale US and CEUS) who was blinded to all the histopathologic results and clinic datum.

2.4 Image analysis

2.4.1 Qualitative analysis

In CEUS and At-PI, the perfusion pattern, perfusion uniformity and color spatial distribution in lesions were recorded and analyzed qualitatively.

In this study, four perfusion patterns were identified: centripetal perfusion (perfusion from periphery to center, ie, colors are changed from red to purple from periphery to center), centrifugal perfusion (perfusion from center to periphery, ie, colors are changed from red to purple from center to periphery), homogeneous perfusion (homogeneous enhancement, ie, colors are distributed homogeneously) and non-perfusion (no color spatial distribution). (Fig. 1)

Fig.1

Perfusion patterns, perfusion uniformity and color spatial distribution for At-PI. Figure 1 shows side-by-side paired displays of GI (left image of each pair) and AI (right image of each pair) of different breast lesions. (a) Centripetal perfusion, uniform perfusion and type 1 in color spatial distribution. (b) Centrifugal perfusion, uniform perfusion and type 3 in color spatial distribution. (c) Homogeneous perfusion, non-uniform perfusion and type 2 in color spatial distribution. (d) Non-perfusion in perfusion pattern, non-perfusion in perfusion uniformity and type 4 in color spatial distribution.

The perfusion uniformity was classified into three types: uniform perfusion (colors are distributed homogeneously within the lesion), non-uniform perfusion (there are some no-enhancement regions within the lesion) and non-perfusion. (Fig. 1)

The color spatial distribution was identified: type1 (lesions are mainly red, orange or yellow), type2 (colors are distributed homogeneously), type3 (lesions are mainly green, blue or purple) and type4 (no color spatial distribution). (Fig. 1)

2.4.2 Quantitative analysis

In CEUS and At-PI, with the dual-image contrast mode, maximal diameters and areas of lesions were measured and MDR_AI/GI and AR_AI/GI were calculated and recorded respectively.

2.5 The interobserver reproducibility

Qualitative and quantitative features in CEUS and At-PI were evaluated respectively and independently by one beginner and one expert (he was the expert who performed At-PI and qualitative and quantitative analysis) who were both blinded to histopathologic results and clinic datum. These features included perfusion patterns, perfusion uniformity, color spatial distribution, MDR_AI/GI, AR_AI/GI, the maximal diameter ratio of the lesion in CI and that in GI (MDR_CI/GI) and the area ratio of the lesion in CI and that in GI (AR_CI/GI). The interobserver reproducibility was evaluated.

2.6 The intraobserver reproducibility

The beginner and the expert separately and independently evaluated the same features in At-PI after a two-week hiatus. These features included perfusion patterns, perfusion uniformity, color spatial distribution, MDR_AI/GI and AR_AI/GI.

2.7 Statistical analysis

SPSS version 20.0 (SPSS, Inc., Chicago, IL, USA) was used for data analysis. Intraclass Correlation Coefficient (ICC) was calculated to evaluate the reproducibility of MDR_AI/GI, AR_AI/GI, MDR_CI/GI and AR_CI/GI. The Kappa value was calculated to evaluate the agreement of perfusion patterns, perfusion uniformity and color spatial distribution. The degree of agreement was set as the following [9]: 0–0.20, slight agreement; 0.21–0.40, poor agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; 0.81–1.00, excellent agreement. The means±SD were used to express the continuous data. The Mann-Whitney U Test was used to compare continuous data and Kruskal-Wallis test was used to compare categorical data. The receiver operating characteristic curves (ROCs) for MDR_AI/GI and AR_AI/GI were plotted. And by maximizing Youden’s index, the optimal cutoff point was obtained. Then, the diagnostic performance of MDR_AI/GI and AR_AI/GI was calculated form the aspect of sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV) and the area under the receiver operating characteristic curve (AUC) with the 95% confidence interval. The z test was used to compare the AUCs of MDR_AI/GI and AR_AI/GI. The t test was used to compare the Kappa values for the intraobserver reproducibility. In all tests, P < 0.05 was accepted to be statistically significant.

3 Results

3.1 Histopathologic diagnosis

All the histopathological results were obtained by the surgical excision or the percutaneous biopsy. In total of 184 breast lesions, 116 were benign and 68 were malignant. Detailed histopathologic diagnosis were shown in Table 1.

Table 1
Histopathologic diagnosis of 184 breast lesions

Benign lesions N Malignant lesions N

Fibroadenoma 58 Invasive ductal carcinoma 54

Adenosis 31 Ductal carcinoma in situ 6

Fibroadenoma with adenosis 14 Lobular carcinoma in situ 2

Intraductal papilloma 4 Mucinous carcinoma 5

Acute suppurative mastitis 2 Metaplastic carcinoma with squamous cell component 1

Granulomatous inflammation 2

Cyst with hemorrhage 4

Mammary neurilemmoma 1

Total 116 Total 68

Benign lesions	N	Malignant lesions	N
Fibroadenoma	58	Invasive ductal carcinoma	54
Adenosis	31	Ductal carcinoma in situ	6
Fibroadenoma with adenosis	14	Lobular carcinoma in situ	2
Intraductal papilloma	4	Mucinous carcinoma	5
Acute suppurative mastitis	2	Metaplastic carcinoma with squamous cell component	1
Granulomatous inflammation	2
Cyst with hemorrhage	4
Mammary neurilemmoma	1
Total	116	Total	68

3.2 The objectivity and reproducibility of parametric imaging

3.2.1 The interobserve reproducibility

Evaluated by the expert and the beginner, the agreements of perfusion patterns, the perfusion uniformity, the ratio of the maximal diameter and the ratio of the area in AI were better than those in CI. The agreement of perfusion patterns in CEUS was moderate (Kappa 0.586), while that in AI was excellent (Kappa 0.920), with statistically significant difference between them (t = 5.22, P < 0.05). Between the two physicians, both good agreements of perfusion uniformity in CI and AI were seen respectively with Kappa values of 0.767 and 0.791. And there was no statistically significant difference in the perfusion uniformity between them (t = 0.375, P > 0.05). The ICC values of MDR_CI/GI and MDR_AI/GI were 0.833 and 0.934, respectively. Similarly, the ICC values of AR_CI/GI and AR_AI/GI were 0.878 and 0.945. The interobserver reproducibility between the expert and the beginner was good. (Table 2)

Table 2
The interobserver reproducibility in features in CEUS and At-PI

Feature Kappa ICC P value

CEUS

Perfusion pattern 0.586 – <0.001

Perfusion uniformity 0.767 – <0.001

MDR_CI/GI – 0.833 –

AR_CI/GI – 0.878 –

At-PI

Perfusion pattern 0.920 – <0.001

Perfusion uniformity 0.791 – <0.001

MDR_AI/GI – 0.934 –

AR_AI/GI – 0.945 –

Feature	Kappa	ICC	P value
CEUS
Perfusion pattern	0.586	–	<0.001
Perfusion uniformity	0.767	–	<0.001
MDR_CI/GI	–	0.833	–
AR_CI/GI	–	0.878	–
At-PI
Perfusion pattern	0.920	–	<0.001
Perfusion uniformity	0.791	–	<0.001
MDR_AI/GI	–	0.934	–
AR_AI/GI	–	0.945	–

3.2.2 The intraobserver reproducibility

For perfusion patterns, perfusion uniformity and color spatial distribution in AI, the intraobserver Kappa values for the beginner were 0.901, 0.883 and 0.913, while those for the expert were 0.931, 0.893 and 0.928. Similarly, for the intraobserver reproducibility of MDR_AI/GI and AR_AI/GI, the ICC values for the beginner were 0.983 and 0.991, while those for the expert were 0.994 and 0.995. These results suggested good intraobserver reproducibility. (Table 3)

Table 3
The intraobserver reproducibility of features in At-PI

Features The beginner The expert

Kappa ICC P value Kappa ICC P value

Perfusion pattern 0.901 – <0.001 0.931 – <0.001

Perfusion uniformity 0.883 – <0.001 0.893 – <0.001

Color spatial distribution 0.913 – <0.001 0.928 – <0.001

MDR_AI/GI – 0.983 – – 0.994 –

AR_AI/GI – 0.991 – – 0.995 –

Features	The beginner	The expert
Perfusion pattern	0.901	–	<0.001	0.931	–	<0.001
Perfusion uniformity	0.883	–	<0.001	0.893	–	<0.001
Color spatial distribution	0.913	–	<0.001	0.928	–	<0.001
MDR_AI/GI	–	0.983	–	–	0.994	–
AR_AI/GI	–	0.991	–	–	0.995	–

3.3 The comparisons of qualitative features in At-PI between benign and malignant breast lesions

Totally, 64 cases of benign lesions (64/116, 55.2%) and 50 cases of malignant lesions (50/68, 73.5%) exhibited centripetal perfusion. 12 non-perfusion lesions were all benign.The perfusion patterns were significantly different between benign and malignant lesions (P = 0.009).

Many benign lesions (54/116, 46.6%) exhibited uniform perfusion and most malignant lesions (36/68, 52.9%) exhibited non-uniform perfusion. However, there was no signficant difference in perfusion uniformity between benign and malignant lesions (P = 0.448).

Colors were commonly distributed homogeneously (type 1) in the 41.4% of benign lesions (48/116). And 25.9% of benign lesions (30/116) mainly performed a mixed green, blue or purple distribution (type 3). However, 64.7% of malignant lesions (44/68) mainly performed a mixed red, orange or yellow distribution (type 1). And 29.4% of them (20/68) were homogeneous in color spatial distribution (type 2). Finally, the difference in color spatial distribution was statistically significant (P < 0.001). The detailed comparisons of qualitative features were shown in Table 4.

Table 4
The comparisons of qualitative features in At-PI between benign and malignant breast lesions

Qualitative features Benign (n = 116) Malignant (n = 68) Z P value

Perfusion pattern 6.875 0.009

centripetal 64 (55.20) 50 (73.50)

centrifugal 22 (19.00) 8 (11.80)

homogeneous 18 (15.50) 10 (14.70)

non-perfusion 12 (10.30) 0 (0.00)

Perfusion uniformity 0.576 0.448

uniform 54 (46.60) 32 (47.10)

non-uniform 50 (43.10) 36 (52.90)

non-perfusion 12 (10.30) 0 (0.00)

Color spatial distribution 37.641 <0.001

Type 1 26 (22.40) 44 (64.70)

Type 2 48 (41.40) 20 (29.40)

Type 3 30 (25.90) 4 (5.90)

Type 4 12 (6.50) 0 (0.00)

Qualitative features	Benign (n = 116)	Malignant (n = 68)	Z	P value
Perfusion pattern			6.875	0.009
centripetal	64 (55.20)	50 (73.50)
centrifugal	22 (19.00)	8 (11.80)
homogeneous	18 (15.50)	10 (14.70)
non-perfusion	12 (10.30)	0 (0.00)
Perfusion uniformity			0.576	0.448
uniform	54 (46.60)	32 (47.10)
non-uniform	50 (43.10)	36 (52.90)
non-perfusion	12 (10.30)	0 (0.00)
Color spatial distribution			37.641	<0.001
Type 1	26 (22.40)	44 (64.70)
Type 2	48 (41.40)	20 (29.40)
Type 3	30 (25.90)	4 (5.90)
Type 4	12 (6.50)	0 (0.00)

Note: Data in parentheses are percentages. Type 1. Lesions are mainly red, orange or yellow. Type 2. Colors are distributed homogeneously. Type 3. Lesions are mainly green, blue or purple. Type 4. Non-perfusion and no color spatial distribution.

3.4 The comparisons of quantitative features in At-PI between benign and malignant breast lesions

Compared to the sizes of lesions in GI, maximal diameters of 54 malignant lesions (54/68, 79.4%) and areas in 58 (58/68, 85.3%) were increased in AI. The mean values of MDR_AI/GI and AR_AI/GI in malignant lesions were both higher than those in benign ones (MDR_AI/GI, 1.53±0.45 versus 1.04±0.15) (AR_AI/GI, 2.00±0.99 versus 1.03±0.21). And there were statistically significnat differences between benign and malignant lesions (P_MDRAI/GI<0.001, P_ARAI/GI<0.001). (Table 5)

Table 5
The comparisons of quantitative features in At-PI and diagnostic performance of At-PI for breast lesions

Features Benign Malignant Z P value Cut-off value Sensitivity (%) Specificity (%) Accuracy (%) PPV (%) NPV (%) AUC (95% confidence interval)

MDR_AI/GI 1.04±0.15 1.53±0.45 –8.927 <0.001 1.125 84.48 (98/116) 88.24 (60/68) 85.87 (158/184) 92.45 (98/106) 76.92 (60/78) 0.895 (0.840–0.949)

AR_AI/GI 1.03±0.21 2.00±0.99 –10.269 <0.001 1.21 93.1 (108/116) 91.18 (62/68) 92.39 (170/184) 94.74 (108/114) 88.57 (62/70) 0.954 (0.925–0.982)

Features	Benign	Malignant	Z	P value	Cut-off value	Sensitivity (%)	Specificity (%)	Accuracy (%)	PPV (%)	NPV (%)	AUC (95% confidence interval)
MDR_AI/GI	1.04±0.15	1.53±0.45	–8.927	<0.001	1.125	84.48 (98/116)	88.24 (60/68)	85.87 (158/184)	92.45 (98/106)	76.92 (60/78)	0.895 (0.840–0.949)
AR_AI/GI	1.03±0.21	2.00±0.99	–10.269	<0.001	1.21	93.1 (108/116)	91.18 (62/68)	92.39 (170/184)	94.74 (108/114)	88.57 (62/70)	0.954 (0.925–0.982)

Note: Data in parentheses are percentages.

The cutoff values for MDR_AI/GI and AR_AI/GI and AUCs were calculated by ROC. The cutoff values for MDR_AI/GI and AR_AI/GI were 1.125 and 1.21, respectively. The AUCs for MDR_AI/GI and AR_AI/GI were 0.895 and 0.954, respectively, with no significant difference between them (Z = 1.84, P > 0.05). A criterion of malignancy was considered as the presence of MDR_AI/GI and AR_AI/GI exceeding the cutoff values. Then the diagnostic performance of At-PI for the differentiation between benign and malignant breast lesions was evaluated. By the thresholds of 1.125 for MDR_AI/GI and 1.21 for AR_AI/GI, the sensitivity, specificity, accuracy, PPV and NPV of At-PI were 84.48%, 88.24%, 85.57%, 92.45% and 76.92%, respectively, for MDR_AI/GI and 93.10%, 91.18%, 92.39%, 94.74% and 88.57%, respectively, for AR_AI/GI. (Fig. 2, Table 5)

Fig.2

The ROC curve for MDR_AI/GI and AR_AI/GI.

With 1.21 being the optimal cutoff for AR_AI/GI, 85.3% (58/68) of the malignant lesions had AR_AI/GI above the threshold, while 14.71% (10/68) of them had AR_AI/GI below the threshold. 2 of the 10 malignant lesions were lobular carcinomas in situ. 2 were ductal carcinomas in situ. And 6 were invasive ductal carcinomas (2 cases histologic grade I, 4 cases histologic grade II). In addition, it was observed that 6.9% (8/116) of the benign lesions had area increase above the 1.21 threshold for AR_AI/GI. In the 8 cases of the benign lesions, 2 were acute suppurative mastitis, 2 were granulomatous inflammation, 2 were intraductal papillomas, and 2 were fibroadenomas with adenosis. With 1.125 being the optimal cutoff for MDR_AI/GI, 20.6% (14/68) of the malignant lesions had MDR_AI/GI below the threshold. The 14 cases of malignancy were misdiagnosed, including 10 cases that had AR_AI/GI below the threshold and 4 invasive ductal carcinomas (histology grade I). Otherwise, 12.1% (14/116) of the benign lesions had MDR_AI/GI above the threshold, which included 8 cases that had AR_AI/GI above the threshold, 4 fibroadenomas with adenoses and 2 adenoses.

4 Discussion

CEUS reveals tumor microcirculation and visualizes tiny blood vessels in real-time. However, in contrast-enhanced ultrasound images, the degree of enhancement of lesions is represented monochromatically, which makes it difficult for some physicians to identify the lesion boundary accurately from breast parenchyma, especially in CI alone. Whereas, in At-PI, contrasting colors are coded according to different arrival times of the contrast agent, which visualizes the differences of arrival times and reconstructs vessel structure objectively and intuitively. Moreover, At-PI can clearly exhibit homogeneity or heterogeneity of tumors and improve the visibility of CEUS [10 –13]. In our study, we evaluated the intraobserver and interobserver agreements of perfusion patterns, perfusion uniformity, color spatial distribution, MDR_CI/GI, AR_CI/GI, MDR_AI/GI and AR_AI/GI in AI and CI. We found that the agreements of above features evaluated by the expert and the beginner in AI were better than those in CI, and the values of Kappa and ICC obtained by the same physician in AI were nearly identical. Our results suggested that At-PI displayed good intraobserver and interobserver reproducibility in the differentiation between benign and malignant breast lesions. And there were less operator dependence, less observer bias and higher reliability for At-PI. At-PI can be a promising method for the beginners to improve the diagnostic performance for breast lesions.

Meanwhile, our study found that At-PI was able to solve the problem that CEUS could not sufficiently suppress the signal from echogenic tissue surrounding breast lesions, such as fascia. In CI, fascia is hyperechoic of which signal can not be suffificiently suppressed, while in AI fascia is non-perfused and displays the background color (Fig. 3). For most beginners, it is difficult to distinguish fascia and enhanced lesions, especially in CI alone. Therefore, At-PI is a more accurate and objective imaging technology to identify the lesion boundary from breast parenchyma.

Fig.3

The GI, CI and AI for the breast lesion and fascias (The white arrows point to fascias). (a) The lesion boundary and fascias are shown in GI. The lesion is hypoechoic and fascias are hyperechoic. (b) The lesion boundary and fascias are shown in CI. The lesion is characterized as hyper-enhanced and fascias seem to be also hyper-enhanced, which is difficult for most beginners to accurately identify the lesion boundary, especially in CI. (c) The lesion boundary and fascias are shown in AI. The lesion performs a mixed color distribution. Fascias are hyperechoic but non-perfused.

In our study, the percentage of malignant lesions presenting centripetal perfusion (50/68, 73.5%) was higher than that of benign lesions (64/116, 55.2%) in AI. Many benign lesions exhibited uniform perfusion (54/116, 46.6%) and most malignant lesions exhibited non-uniform perfusion (36/68, 52.9%), which was in accordance with pathologic characteristics of malignancy. In malignant tumors, the vessel spatial distribution is unbalanced and the vessel heterogeneity is more prominent. Malignant lesion vessels are characterised by caliber fluctuations, various and non-unniform vessel thickness, disordered distribution and arteriovenous shunts. Large amounts of vascular endothelial growth factor is produced in the periphery of lesions, which makes a higher peripheral vessel-to-central vessel ratio [14 –16]. Namely, the degree of vascularization in the periphery is higher than in the centre. Meanwhile, the growth rate of malignant cells is higher than the formation of tiny blood vessels, and metabolism is active within malignant tumors. Because of the relatively insufficiency of oxygen and nutrition, the centre of malignant tumors is easier to become hypoxic and necrotic [17 –19]. Otherwise, Liu et al. [20] primarily attributed blood perfusion defect to desmoplastic stroma, low cellularity, dilated duct, degeneration, fibrosis or necrosis. Conversely, the vessels of benign lesions are mostly originated from normal breast vessels. These vessels are uniform in thickness and distribution [21, 22].

Furthermore, the color spatial distribution in lesions is a unique feature in At-PI, which provides more diagnostic information for differentiating benign and malignant breast lesions. Since the same interval is set in the graduated color scale, the area percent of different colors in the whole lesion represents the rate of blood perfusion. In our study, 41.4% of the benign lesions (48/116) performed a homogeneous color distribution (type2), and 64.7% of the malignant lesions (44/68) performed a mixed red, orange or yellow distribution (type 1). The percentage of type 1 in malignant lesions was notably higher than in benign ones. Our results suggested that compared to benign lesions, most malignant ones showed faster perfusion, which is consistent with previous studies [19].

Several studies have demonstrated that range increase after enhancement was considered as a malignant feature [1 –4]. Notwithstanding, so far, there has been no uniform diagnostic criteria to differentiate benign and malignant tumors [23, 24]. Considering MDR_AI/GI and AR_AI/GI as new features, our study demonstrated that the increases of the maximal diameter and the area were efficient indicators for the differentiation between benign and malignant lesions. What’s more, by using the thresholds of 1.125 for MDR_AI/GI, 79.4% (54/68) of the malignant lesions had MDR_AI/GI above the threshold and 87.9% (102/116) of the benign ones had MDR_AI/GI below the threshold. Meanwhile, the diagnostic sensitivity, specificity, accuracy, PPV and NPV of At-PI for MDR_AI/GI were 84.48%, 88.24%, 85.57%, 92.45% and 76.92%, respectively. By using the thresholds of 1.21 for AR_AI/GI, 85.3% (58/68) of the malignant lesions had AR_AI/GI above the threshold and 93.1% (108/116) of the benign ones had MDR_AI/GI below the threshold. And the diagnostic sensitivity, specificity, accuracy, PPV and NPV of At-PI for AR_AI/GI were 93.10%, 91.18%, 92.39%, 94.74% and 88.57%, respectively. Our study also showed that the AUC of AR_AI/GI was larger than that of MDR_AI/GI (0.954 vs 0.895) and they were not statistically different from one another (Z = 1.84, P > 0.05). Therefore, in clinical practice, MDR_AI/GI and AR_AI/GI are both good and efficient features and the use of MDR_AI/GI may be more convenient for the diagnosis for breast lesions with high sensitivity and specificity.

In the most malignant lesions that had MDR_AI/GI or AR_AI/GI above the thresholds, it was observed that the stromal response was seen around them, especially invasive ductal carcinomas. Therefore, we supposed that the range increase for malignant lesions after enhancement was associated with the stromal response. The carcinoma-associated fibroblast (CAF) is an important component of the stroma of breast cancers and the key element in tumor development and metastasis, which includes the fibroblasts and the myofibroblasts. CAFs can promote tumor stromal response. Tumor cells trigger the activation of the stroma. Then, the abnormal growth of fibrotic tissue, known as the desmoplastic reaction, will be seen around tumors [25], which may be a defensive mechanism to isolate the tumor [26]. Pathologically, the desmoplastic reaction in the tumor stroma can help to diagnose invasion [27]. It is a hallmark of malignancy that large numbers of myofibroblasts exist in the stroma [26]. In this study, we observed that a number of the fibroblasts existed around the tumor cells in most invasive breast cancers. In addition, dilated vessels were also observed (Fig. 4). While, few fibroblasts were observed in most ductal carcinomas in situ, lobular carcinomas in situ and benign lesions, and the desmoplastic reaction was not commonly seen (Fig. 4). Those may be the pathological basis for the range increase of most malignancy after enhancement.

Fig.4

Histopathologic sections for malignant and benign breast lesions. (a) Histopathologic examination confirms the diagnosis of an invasive ductal carcinoma (hematoxylin-eosin stain). The asterisk represents the tumor cells. (b) The figure shows the amplified histopathologic section in the black box in Fig. 4-a. A number of the fibroblasts exist around the tumor cells and the dilated vessels can be seen. The asterisk represents the tumor cells. The black arrow points to the dilated vessel. The white arrows point to the fibroblasts. (c) Histopathologic examination confirms the diagnosis of a fibroadenoma (hematoxylin-eosin stain). F represents the tumor. (d) The figure shows the amplified histopathologic section in the black box in Fig. 4-c. The boundary of the tumor is clear. Few fibroblasts can be seen around the tumor. F represents the tumor. The black line shows the boundary of the tumor.

In this study, with 1.21 being the optimal cutoff for AR_AI/GI, 10 cases (10/68, 14.7%) of the malignancy had AR_AI/GI below the threshold. 2 of them were lobular carcinomas in situ, 2 were ductal carcinomas in situ, and 6 were invasive ductal carcinomas (2 cases histologic grade I, 4 cases histologic grade II). Because the stromal response was not commonly seen in most ductal carcinomas in situ and lobular carcinomas in situ, those lesions had AR_AI/GI below the threshold. In other 6 cases of invasive ductal carcinomas, there were 2 cases with histologic grade I and 4 cases with histologic grade II. We supposed that with the higher histologic grade and the richer blood perfusion, tumors would show more atypical malignant features [28]. Inflammatory responses can be characterised by inflammatory cell infiltration and local vascular dilation. And inflammation often extends to surrounding adjacent perilobular and interlobular tissue, which might be the reason why inflammatory lesions often display high enhancement and range increase. Fibroadenomas have capsules and margins are commonly clear. Fibroadenomas are not commonly seen range increase after CEUS. While adenoses have no capsules and sometimes there is not a clear boundary between adenoses and surrounding breast parenchyma after CEUS. In this study, the margins of 2 cases of fibroadenomas with adenoses were not clear and area increase were seen after CEUS.

With 1.125 being the cutoff value for MDR_AI/GI, 14 cases (14/68, 20.6%) of the malignancy were misdiagnosed, including 10 cases that had AR_AI/GI below the threshold and 4 invasive ductal carcinomas (histologic grade I). Otherwise, 12.1% (14/116) of the benign lesions had MDR_AI/GI above the threshold, which included 8 cases that had AR_AI/GI above the threshold, 4 fibroadenomas with adenoses and 2 adenoses.Our study found that with MDR_AI/GI being the diagnostic indicator, there were more misdiagnosed cases. We supposed that MDR_AI/GI ignored the growth and infiltration for breast lesions in the depth, while AR_AI/GI took breast lesions with both the width and the depth into consideration and AR_AI/GI could evaluate the range of lesions more accurately. Moreover, several studies have reported that elastography might add diagnostic value to CEUS in the diagnosis of breast lesions [29 –32]. Therefore, the use of elastography may improve the diagnostic performance in our study.

There are still some limitations in this study. Firstly, the sample size was small and the pathological subtypes were simple. The sample sizes of benign and malignant lesions were not sufficiently balanced. Secondly, CEUS might only show the tumor margin vessels in BI-RADS categories 4-5. Thirdly, parametric images are easily influenced by the quality of contrast-enhanced ultrasound images. The stability of probes and breath and heartbeat of patients can affect the accuracy of the color spacial distributian. So that the quality control of contrast-enhanced images is crucial.

5 Conclusions

At-PI is more objective and intuitive to reconstruct vessel structure, accurately assess the lesion size and improve the visibility of CEUS. What’s more, At-PI displays good intraobserver and interobserver reproducibility and less observer bias, which is an efficient and powful method to straightforwardly visualize the real-time hemodynamics. Meanwhile, MDR_AI/GI and AR_AI/GI could be efficient features for the differential diagnosis of breast lesions.

Conflicts of interest

We confirm that there are no conflicts of interest to declare.

Footnotes

Acknowledgments

The authors acknowledge the support from NSFC/China (81671687).

References

Wang

, Fan

, Zhao

, Zhang

, et al. Qualitative, quantitative and combination score systems in differential diagnosis of breast lesions by contrast-enhanced ultrasound. Eur J Radiol. 2016;85(1):48–54.

Luo

, Chen

, Yue

, Zhou

, Lan

, et al. Predictive model for contrast-enhanced ultrasound of the breast: Is it feasible in malignant risk assessment of breast imaging reporting and data system 4 lesions? World J Radiol. 2016;8(6):600–9.

Quan

, Hong

, Zhang

, Mei

, You

, Huang

. The clinical role of contrast enhanced ultrasound in differential diagnosis of BI-RADS 4 breast lesions. Clin Hemorheol Microcirc. 2019;72(3):293–303.

Shao

, Li

, Yao

, Li

, Wu

. Incorporation of contrast-enhanced ultrasound in the differential diagnosis for breast lesions with inconsistent results on mammography and conventional ultrasound. Clin Hemorheol Microcirc. 2019 Dec 19. doi: 10.3233/CH-190729.

Jiang

, Liu

, Zhu

, Sun

, Chang

. Breast tumor size assessment: comparison of conventional ultrasound and contrast-enhanced ultrasound. Ultrasound Med Biol. 2007;33(12):1873–81.

Zheng

, Xu

, Liu

, Wang

, Zhang

, Guo

, et al. Parametric imaging with contrast-enhanced ultrasound: usefulness for characterization of dynamic effects of microvascularization for hepatocellular carcinoma and focal nodular hyperplasia. Clin Hemorheol Microcirc. 2013;55(3):375–89.

Wakui

, Takayama

, Kanekawa

, Ichimori

, Otsuka

, Shinohara

, et al. Usefulness of arrival time parametric imaging in evaluating the degree of liver disease progression in chronic hepatitis C infection. J Ultrasound Med. 2012;31(3):373–82.

Noro

, Nakamura

, Hirai

, Haga

, Kobayashi

, Hayashi

, et al. Impact of parametric imaging on contrast-enhanced ultrasound of breast cancer. J Med Ultrason. 2016;43(2):227–35.

Landis

, Koch

. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159–74.

10.

Wakui

, Nagai

, Matsukiyo

, Ogino

, Matsui

, Mukozu

, et al. Effect of Hepatic Inflammation in Chronic Hepatitis C Infection on Fibrosis Assessment by Arrival Time Parametric Imaging. Ultrasound Q. 2018;34(3):128–32.

11.

Yuan

, Quan

, Yunxiao

, Jian

, Zhu

, Liping

. Diagnostic value of contrast-enhanced ultrasound parametric imaging in breast tumors. J Breast Cancer. 2013;16(2):208–13.

12.

Wakui

, Takayama

, Kamiyama

, Kobayashi

, Matsui

, Matsukiyo

, et al. Arrival time parametric imaging using Sonazoid-enhanced ultrasonography is useful for the detection of spoke-wheel patterns of focal nodular hyperplasia smaller than 3cm. Exp Ther Med. 2013;5(6):1551–4.

13.

Yin

, Cui

, Fan

, Yang

, Yan

. Diagnostic Value of Arrival Time Parametric Imaging Using Contrast-Enhanced Ultrasonography in Superficial Enlarged Lymph Nodes. J Ultrasound Med. 2019;38(5):1287–98.

14.

, Wu

, Chen

, Gu

. Application of contrast-enhanced ultrasound in the diagnosis of small breast lesions. Clin Hemorheol Microcirc. 2018;70(3):291–300.

15.

, Li

, Zhu

, Fu

, Liu

. Association between enhancement patterns and parameters of contrast-enhanced ultrasound and microvessel distribution in breast cancer. Oncol Lett. 2018;15(4):5643–9.

16.

Jung

, Jungius

, Rupp

, Gallegos

, Ritter

, Lenhart

, et al. Contrast enhanced harmonic ultrasound for differentiating breast tumors - first results. Clin Hemorheol Microcirc. 2005;33(2):109–20.

17.

Cao

, Bao

, Zhu

, Wang

, Sun

, Wang

, et al. Contrast-enhanced ultrasound characteristics of breast cancer: correlation with prognostic factors. Ultrasound Med Biol. 2014;40(1):11–7.

18.

Metz

, Daldrup-Unk

, Richter

, Rath

, Ebert

, Settles

, et al. Detection and quantification of breast tumor necrosis with MR imaging: value of the necrosis-avid contrast agent Gadophrin-3. Acad Radiol. 2003;10(5):484–90.

19.

Vraka

, Panourgias

, Sifakis

, Koureas

, Galanis

, Dellaportas

, et al. Correlation Between Contrast-enhanced Ultrasound Characteristics (Qualitative and Quantitative) and Pathological Prognostic Factors in Breast Cancer. In Vivo. 2018;32(4):945–54.

20.

Zhao

, Liu

, Wei

, Xu

, Wu

, Xu

, et al. Contrast-Enhanced Ultrasonography Features of Breast Malignancies with Different Sizes: Correlation with Prognostic Factors. Biomed Res Int. 2015;2015:613–831.

21.

Ozdemir

, Ozdemir

, Maral

, Konus

, Yucel

, Isik

. Differential diagnosis of solid breast lesions: contribution of Doppler studies to mammography and gray scale imaging. J Ultrasound Med. 2001;20(10):1091–1101, 1102.

22.

Zhao

, Xu

, Ouyang

, Chen

, Dong

, Huihua

. Contrast-enhanced ultrasound is helpful in the differentiation of malignant and benign breast lesions. Eur J Radiol. 2010;73(2):288–93.

23.

, Gong

, Ling

, Du

, Su

, Wang

, et al. Diagnostic performance of contrast-enhanced ultrasound and enhanced magnetic resonance for breast nodules. J Biomed Res. 2018;32(3):198–207.

24.

Quan

, Hong

, Zhang

, Mei

, You

, Huang

. The clinical role of contrast enhanced ultrasound in differential diagnosis of BI-RADS 4 breast disease. Clin Hemorheol Microcirc. 2019;72(3):293–303.

25.

Martinez-Outschoorn

, Pavlides

, Whitaker-Menezes

, Daumer

, Milliman

, Chiavarina

, et al. Tumor cells induce the cancer associated fibroblast phenotype via caveolin-1 degradation: Implications for breast cancer and DCIS therapy with autophagy inhibitors. Cell Cycle. 2010;9(12):2423–33.

26.

Schäfer

, Werner

. Cancer as an overhealing wound: an old hypothesis revisited. Nature Reviews Molecular Cell Biology. 2008;9(8):628–38.

27.

Bauer

, Su

, Casper

, He

, Rehrauer

, Friedl

. Heterogeneity of gene expression in stromal fibroblasts of human breast carcinomas and normal breast. Oncogene. 2010;29(12):1732–40.

28.

Wang

, Xu

, Wang

, Grant

. Contrast-enhanced ultrasonographic findings of different histopathologic types of breast cancer. Acta Radiol. 2011;52(3):248–55.

29.

Zhang

, Wan

, Wu

, Fang

. Value of conventional ultrasound, ultrasound elasticity imaging, and acoustic radiation force impulse elastography for prediction of malignancy in breast lesions. Clin Hemorheol Microcirc. 2019 Oct 30. doi: 10.3233/CH-180527

30.

Georgieva

, Prantl

, Utpatel

, Wiesinger

, Stroszczynski

, Jung

. Diagnostic performance of ultrasound strain elastography for differentiation of malignant breast lesions. Clin Hemorheol Microcirc. 2019;71(2):237–47.

31.

, Yan

, Liu

, Tang

, Zhou

, Wu

, Xue

, Shen

, Tang

, Chen

, Xue

, Li

, Guo

, Wang

, Li

, Yan

, Li

, Wang

, Zhang

, Wu

, Yu

, Zhou

. Guidelines and recommendations on the clinical use of shear wave elastography for evaluating thyroid nodule1. Clin Hemorheol Microcirc. 2019;72(1):39–60.

32.

, Zhang

, Xiang

, Zhang

, Xu

, Liu

, Tang

, Zhao

, Wu

. The efficacy of added shear wave elastography (SWE) in breast screening for women with inconsistent mammography and conventional ultrasounds (US). Clin Hemorheol Microcirc. 2019;71(1):83–94.