Quantitative analysis of shear wave elastic heterogeneity for prediction of lymphovascular invasion in breast cancer

Abstract

Objective:

To evaluate the correlation between elastic heterogeneity (EH) and lymphovascular invasion (LVI) in breast cancers and assess the clinical value of using EH to predict LVI pre-operatively.

Methods:

This retrospective study consisted of 376 patients with breast cancers that had undergone shear wave elastography (SWE) with virtual touch tissue imaging quantification between June 2017 and June 2018. The EH was determined as the difference between the averaged three highest and three lowest shear wave value. Clinicalpathological parameters including histological type and grades, LVI, axillary lymph node status and molecular markers (estrogen receptor, progesterone receptor, human epidermal growth factor receptor 2 and Ki-67) were reviewed and recorded. Relationship EH and clinicalpathological parameters was investigated respectively. The diagnostic performance of EH in distinguishing LVI or not was analyzed.

Results:

At multivariate regression analysis, only EH (p = 0.017) was positively correlated with LVI in all tumors. EH (p = 0.003) and Ki-67 (p = 0.025) were positively correlated with LVI in tumors ≤ 2 cm. None of clinicalpathological parameters were correlated with LVI in tumors > 2 cm (p > 0.05 for all). Using EH to predict LVI in tumors ≤ 2 cm, the sensitivity and negative predictive value were 93 and 89% respectively.

Conclusion:

EH has the potential to be served as an imaging biomarker to predict LVI in breast cancer especially for tumors ≤ 2 cm.

Advances in knowledge:

There was no association between LVI and other most commonly used elastic features such as SWV_mean and SWV_max. Elastic heterogeneity is an independent predictor of LVI, so it can provide additional prognostic information for routine preoperative breast cancer assessment.

For tumors ≤ 2cm, using EH value higher than 1.36 m/s to predict LVI involvement, the sensitivity and negative predictive value can reach to 93% and 89%, respectively, suggesting that breast cancer with negative EH value was more likely to be absent of LVI.

Introduction

Recurrence and metastasis are the major causes of breast cancer-related death.¹ Lymphovascular invasion (LVI) is a pathological finding of carcinoma in the lymphatic system and vascular system surrounding the invasive tumoral tissues.² Tumor cells which are able to cause LVI are more likely to spread to lymph nodes and to distant sites.³ LVI was an independent and poor prognostic factor of invasive breast cancer.⁴

Pre-operative prediction of LVI could guide the choice of aggressive axillary lymph node dissection (ALND) for sentinel lymph node (SLN) positive patients.⁵ There are 40–60% of SLNs-positive patients had no additional lymph node metastasis (LNM) after ALND.^6,7 Zheng, et al and Viale, et al reported that the incidence of non-SLN involvement were much higher in breast cancer with LVI than in those without LVI.^6,8 Patients who have LVI will be more likely to benefit from ALND. In node-negative patients, LVI was one of predictors of increased risk of locoregional failure.⁹ LVI was associated with higher rates of re-excision after breast conservation therapy.¹⁰ Therefore, the preoperative evaluation of LVI could guide the choice of breast conservation therapy for the patients without clinical LNM. However, it is difficult to preoperatively determine LVI status due to the limited number of samples obtained via core biopsy.

Quantitative apparent diffusion coefficient obtained by preoperative MRI has been used to predict LVI for breast cancer by Mori et al.¹¹ Nevertheless, MRI is not extensively used due to the cost. In contrast, ultrasound examination is convenient and economical, which has become a routine imaging method of pre-operative evaluation for breast cancers.

Shear wave elastography (SWE) have become a complementary tool for conventional ultrasound in the differentiation of breast lesions.¹² Mean elasticity (E_mean) is determined to reflect the mean stiffness of lesion. Recent studies tried to assess the LVI by measurement of E_mean.^13,14 Although LVI was associated with higher E_mean at univariate analysis, no association between LVI and E_mean was found at multivariate analysis.^13,15

Intratumor heterogeneity depicted by imaging can be used to distinguish tumor subtypes, monitor treatment response and predict prognosis.¹⁶ Elastic heterogeneity (EH) refers to the heterogeneity of distribution of stiffness and analyzing EH of breast lesion has been shown to be valuable for distinguishing benign and malignant breast lesions.¹⁷ Virtual touch tissue imaging quantification (VTIQ) is a two-dimensional SWE. It not only has a color-coded map displaying the distribution of shear wave velocity (SWV) and enable us to measure SWV quantitatively, but also provides a quality map to assess the reliability of SWV measurements.¹⁸ In our previous study, we measured three highest and three lowest SWVs on VTIQ and calculated EH by stiffness gradient. The results showed that EH had significantly higher sensitivity than the maximum and mean shear wave velocity (SWV_max and SWV_mean) in distinguishing benign and malignant breast lesions.¹⁹

Therefore, the purpose of our study is to evaluate the correlation of LVI with EH, and assess the role of EH in pre-operatively predicting LVI status of breast cancer.

Methods and materials

Study population

This retrospective study was approved by the Institutional Review Board and patient informed consent was waived. 727 consecutive patients with 727 ultrasound (US) BI-RADS 4–5 breast lesions that underwent breast conventional ultrasound and SWE between June 2017 and June 2018 at our institution (Sun Yat-sen University Cancer Center) were identified. All lesions underwent ultrasound-guided core needle biopsy immediately after SWE examinations. According to the pathological results of biopsy, 313 patients with 313 benign lesions were excluded, and 414 patients with solid breast cancer were investigated. Based on the clinical information and radiologic finding, patients were excluded if: 1. Lesion was >3.5 cm on ultrasound (n = 20) due to the limited width of the transducer; 2. SWE images quality was poor (the whole lesion area was coded yellow or red on quality map) (n = 5); 3. Patients received neoadjuvant chemotherapy (n = 10); 4. Lesions were diagnosed as carcinomas in situ by surgery (n = 3). Finally, a total of 376 women with 376 breast cancer were included in this study (Figure 1).

Figure 1.

Patient selection workflow. BI-RADS = Breast Imaging Reporting and Data System.

Shear wave elastography

Conventional breast ultrasound and SWE were performed by one of the two radiologists (J.H.Z. and Y.N.H.) who had 15 and 3 years of experience, respectively, performing breast ultrasound and at least 2 year of experience performing breast elastography. All images were acquisition by using Siemens S2000 ultrasound system (Siemens Healthineers, Mountain View, CA).

Conventional ultrasound was performed for each lesion by using a 18L linear array transducer. Tumor size was determined as the maximum diameter which was measured on B-mode image.

SWE with VTIQ¹⁸ was performed by using a 9L4 linear transducer along the maximum diameter of the breast lesion. A square sampling region of interest (ROI) was placed on velocity map. The sampling ROI size for each lesion was different and was determined according to lesion size to ensure that the lesion and at least 5 mm of breast tissue adjacent to the lesion could be included. After initiating the acoustic radiation force impulse, a quality map and a velocity map were obtained. The quality map was a 2D color-coded image in which green represented a high quality and reliable SWV measurement, while yellow and red represented an intermediate and low quality SWV measurement that was considered unreliable due to the low shear wave amplitude and unexpected artifacts, i.e. “vertical striped”, movement artifacts and so on.²⁰ Velocity map was a 2D color-coded image where blue represented a low SWV, yellow or green represented an intermediate SWV, and red represented a high SWV. Three times of VTIQ were performed for each lesion. The quality map with the least coded yellow and red areas within lesion, and its corresponding velocity map were selected for analysis.

For each lesion, the measurement of SWV were performed by the one of radiologist (Y.N.H. or J.H.Z.) on each velocity map. After the acquisition of VTIQ maps, maximum value of the scale (ranging from 6.5 to 10 m/s) was adjusted to be slightly (about 0.5 m/s) higher than the highest SWV of the lesion to ensure that the highest SWV of the lesion can be accurately measured. After the scale adjustment, six measurement ROIs with fixed size (about 2 × 2 mm) were placed on the three reddest areas (corresponding to the highest SWV) and three bluest areas (corresponding to the lowest SWV) within or adjacent the lesion if the “stiff rim” (defined as the area around the tumor with abnormally increased elastic stiffness) was presented.²¹ Because different area of breast lesion on the velocity map showed different SWE value, the value of SWE would change as the ROI move over the lesion, which enable us to measure the highest or lowest SWV part of the lesion. The areas coded red or yellow on quality map were avoided on measurements.

In velocity map, the three measurements with highest SWVs were labeled as SWV_max1, SWV_max2, and SWV_max3, and other three measurements with lowest SWVs were labeled as SWV_min1, SWV_min2, and SWV_min3 (Figure 2). SWV_max, SWV_mean and EH were evaluated respectively. The SWV_max was determined as the highest value among SWV_max1, SWV_max2, and SWV_max3, and SWV_mean was the average value of the six measurements. The EH was calculated by using the following formula (Figure 2).¹⁹

Figure 2.

Quantitative measurement of the elastography features. The (a) and (b) were the quality map and velocity map of VTIQ respectively. Six ROIs were placed over three areas with highest SWV (labeled as SWV_max1, SWV_max2, and SWV_max3 respectively; Solid white arrows) and three with lowest SWV (labeled as SWV_min1, SWV_min2 and SWV_min3 respectively; white open arrows) within or surround the tumor on the velocity map (b). The areas coded yellow (Solid black arrows) on quality map (a) were avoided during the whole measurement. For this lesion, SWV_max1, SWV_max2, SWV_max3, SWV_min1, SWV_min2 and SWV_min3 were measured as 9.02 m/s, 9.47 m/s, 8.41 m/s, 4.04 m/s, 3.64 m/s and 3.37 m/s respectively. The SWV_max of this lesion was 9.47 m/s. The SWV_mean and EH value were calculated as 6.33 m/s and 5.29 m/s respectively. SWV, shear wave velocity; VTIQ, virtual Touch tissue imaging quantification.

$${\displaystyle \begin{array}{ll}\mathrm{E}\mathrm{H}=& \mathrm{A}\mathrm{V}\mathrm{E}(\mathrm{S}\mathrm{W}{\mathrm{V}}_{\mathrm{m}\mathrm{a}\mathrm{x}1}+\mathrm{S}\mathrm{W}{\mathrm{V}}_{\mathrm{m}\mathrm{a}\mathrm{x}2}+\mathrm{S}\mathrm{W}{\mathrm{V}}_{\mathrm{m}\mathrm{a}\mathrm{x}3})\\ & -\mathrm{A}\mathrm{V}\mathrm{E}(\mathrm{S}\mathrm{W}{\mathrm{V}}_{\mathrm{m}\mathrm{i}\mathrm{n}1}+\mathrm{S}\mathrm{W}{\mathrm{V}}_{\mathrm{m}\mathrm{i}\mathrm{n}2}+\mathrm{S}\mathrm{W}{\mathrm{V}}_{\mathrm{m}\mathrm{i}\mathrm{n}3}).\end{array}}$$ (1)

Pathological evaluation

Surgical-resection specimen was prepared for the study and all hematoxylin & eosin-stained slides were analyzed by experienced breast pathologists at our hospital (Sun Yat-sen University Cancer Center). Pathological features including histologic type, histological and nuclear grades, molecular subtypes, LVI status, axillary lymph node (ALN) status were obtained from the routine pathological reports. On the basis of immunohistochemical staining, the status of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2) and Ki-67 were analyzed. Fluorescence in situ hybridization was further used to confirm HER2 status when the immunohistochemical result of HER2 demonstrated a 2+ positivity. Depending on the status of ER, PR, HER2, and Ki-67, the cases were classified as one of five molecular subtypes: luminal A, luminal B (HER2-), luminal B (HER2+), HER2 enriched, and basal-like.^22,23

Statistical analysis

Data were analyzed using the SPSS v. 22.0 (SPSS, Inc, Chicago, IL) and MedCalc v. 15.2.2 for Windows (Mariakerke, Belgium). Relationships between the values of three SWE features and prognostic factors were compared by using independent t-test or one-way ANOVA tests for data following normal distribution and the Mann–Whitney U test or Kruskal–Wallis test for non-parametric data. Relationships between LVI status and other clinicopathological parameters in all cases, tumors ≤ 2 cm and >2 cm were investigated using uni- and multivariate logistic regression models respectively. Receiver operating characteristic (ROC) (Delong) was used to evaluate the effectiveness of EH value in distinguishing LVI-positive and negative breast cancers. Based on the maximal Youden index, the optimal threshold was chosen. The area under the ROC curve (AUC), sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated. p-values < 0.05 were considered statistically significant.

Results

The median age of the 376 patients included in this study was 50 years (range 28–83 years). The median size of the lesions was 18 mm (range, 5–35 mm). The median value of EH, SWV_max and SWV_mean were 3.28 m/s (range, 0.21–7.71 m/s), 6.42 m/s (range, 2.33–10 m/s) and 4.22 m/s (range, 1.75–7.77 m/s) respectively.

The EH, SWV_max and SWV_mean values of tumors > 2 cm were significantly higher than those of tumors ≤ 2 cm (p < 0.001 for all). There was significant difference in SWV_max and SWV_mean value among different molecular subtype cancers (p = 0.02 for SWV_max and p = 0.01 for SWV_mean), while there was no significant difference in the EH between cancers with different molecular subtype (p = 0.11). Among the five subtypes of cancers, basal-like cancers showed the lowest SWV_max and SWV_mean values, and luminal B (HER2+) cancers showed the highest SWV_max and SWV_mean values. The EH and SWV_max values of cancers with LVI was higher than those of cancers without LVI (p = 0.016 for EH, p = 0.024 for SWV_max), while there was no significant difference in the SWV_mean between LVI-positive cancer and LVI-negative cancer (p = 0.056). The EH, SWV_max and SWV_mean values of cancers with axillary lymph node metastasis was higher than those of cancers without axillary lymph node metastasis (p = 0.001 for EH, p = 0.008 for SWV_max; p = 0.014 for SWV_mean) (Table 1).

Table 1.

Comparison of EH value, SWV_max, and SWV_mean of malignant lesions with morphological and histopathological features

Clinicopathological features (number of lesions)	EH valuea		SWVmax		SWVmean
	Mean ± SDa (m/s)	p- valueb	Mean ± SDa (m/s)	p-valueb	Mean ± SDa (m/s)	p- valueb
Age		0.93		0.51		0.54
<40 (n = 45)	3.39 ± 1.68		6.81 ± 2.15		4.60 ± 1.47
≥40 (n = 331)	3.32 ± 1.70		6.58 ± 2.11		4.42 ± 1.31
Tumor size (cm)c		<0.001		<0.001		<0.001
≤2 (n = 255)	2.91 ± 1.56		6.09 ± 1.97		4.18 ± 1.27
>2 (n = 121)	4.22 ± 1.64		7.69 ± 1.99		5.00 ± 1.28
Histologic typed		0.19		0.14		0.40
IDC (n = 335)	3.35 ± 1.68		6.63 ± 2.09		4.46 ± 1.33
ILC (n = 10)	2.96 ± 2.25		6.02 ± 2.26		4.13 ± 1.32
Mixed (n = 21)	3.63 ± 1.72		7.11 ± 2.18		4.56 ± 1.32
Other (n = 10)	2.40 ± 1.72		5.45 ± 2.23		3.86 ± 1.31
Histologic gradee		0.85		0.85		0.78
I-II (n = 187)	3.35 ± 1.67		6.68 ± 2.08		4.48 ± 1.30
III (n = 163)	3.39 ± 1.69		6.67 ± 2.11		4.48 ± 1.35
Molecular subtypese*		0.11		0.02		0.01
Luminal A (n = 87)	3.14 ± 1.64		6.38 ± 2.12		4.31 ± 1.38
Luminal B (HER2-) (n = 144)	3.29 ± 1.75		6.56 ± 2.19		4.39 ± 1.32
Luminal B (HER2+) (n = 75)	3.79 ± 1.73		7.28 ± 2.09		4.92 ± 1.37
HER2 enriched (n = 33)	3.43 ± 1.43		6.80 ± 1.74		4.51 ± 1.16
Basal-like (n = 34)	3.02 ± 1.63		5.93 ± 1.80		3.98 ± 1.07
Lymphovascular invasion		0.016		0.024		0.056
Negative (n = 258)	3.19 ± 1.71		6.44 ± 2.15		4.35 ± 1.34
Positive (n = 118)	3.64 ± 1.64		6.97 ± 1.99		4.64 ± 1.28
Axillary lymph node metastasis		0.001		0.008		0.014
Absent (n = 221)	3.09 ± 1.68		6.36 ± 2.04		4.30 ± 1.26
Present (n = 155)	3.67 ± 1.67		6.96 ± 2.17		4.64 ± 1.40

SWV, shear wave velocity.

^aEH value = Elastic heterogeneity value; Data are presented as mean ± SD; SD = standard deviation.

^bThe statistical significance was established at p = 0.05.

^cSize was defined as the maximum diameter on B-mode ultrasound images.

^dIDC = invasive ductal carcinoma; ILC = invasive lobular carcinoma; Mixed type: more than one histologic type in one lesion; Other type included mucinous (n = 4), papillary (n = 1), micropapillary (n = 2), neuroendocrine (n = 2), tubular (n = 1).

^eOnly invasive ductal carcinoma and mixed type with invasive ductal carcinoma were evaluated.

^fIn 3 cases the HER-2 status was unclear because the IHC and initial FISH test were inconclusive and a repeat ISH test had not been done, therefore the pathological subtype of these three cases could not be determined.

In all tumors, LVI status was found to be positively associated with EH (p = 0.018), SWV_max (p = 0.023), SWV_mean (p = 0.051), tumor size (p = 0.059), and Luminal B (HER2-) type (p = 0.047) on univariate regression analysis. There was no association between LVI status and other prognostic factors including histologic type, grade, ER, PR, HER-2 and Ki-67 (p > 0.1) (Table 2). On multivariate regression analysis, only EH value was found to be significantly associated with LVI status (p = 0.017) (Table 3).

Table 2.

Results from univariate regression analysis of clinicopathological variables for lymphovascular invasion in all cases

Clinicopathologic Variables	Total Number	LVI Negative (n.)	LVI Positive (n.)	β coefficientb	P value
Age				−0.0005 ± 0.01	0.97
EH value				0.16 ± 0.07	0.018
SWVmax				0.12 ± 0.05	0.023
SWVmean				0.16 ± 0.08	0.051
Tumor size				0.03 ± 0.02	0.059
Histologic type
IDC	335	227	108	0
ILC	10	10	0	−20.46 ± 12710.13	1.00
Mixed	21	13	8	0.26 ± 0.46	0.58
Other	10	8	2	−0.64 ± 0.80	0.42
Histologic grade
I-II	187	126	61	0
III	163	107	56	0.10 ± 0.23	0.65
ER				0.004 ± 0.003	0.16
PR				0.001 ± 0.003	0.84
HER-2				0.19 ± 0.24	0.44
Ki-67				0.007 ± 0.005	0.20
Molecular subtypes
Luminal A	87	66	21	0
Luminal B (HER2-)	144	91	53	0.61 ± 0.30	0.047
Luminal B (HER2+)	75	49	26	0.51 ± 0.35	0.14
HER2 enriched	33	22	11	0.45 ± 0.45	0.31
Basal-like	34	28	6	−0.40 ± 0.52	0.44

EH, elastic heterogeneity; SWV, shear wave velocity.

^aData are expressed as estimates ± standard error.

Table 3.

Results from multivariate regression analysis of clinicopathologic variables and lymphovascular invasion in all cases

Clinicopathological variables	β coefficient	p-value
EH value	0.16 ± 0.07	0.017
SWVmax	0.12 ± 0.05	0.79
SWVmean	0.16 ± 0.08	0.99
Tumor size	0.03 ± 0.02	0.30
Molecular subtypes
Luminal A	0	0.15
Luminal B (HER2-)	0.61 ± 0.30	0.06
Luminal B (HER2+)	0.51 ± 0.35	0.71
HER2 enriched	0.45 ± 0.45	0.82
Basal-like	−0.40 ± 0.52	0.09

SWV, shear wave velocity.

For tumors ≤ 2 cm, LVI status was found to be positively associated with EH (p = 0.003), SWV_max (p = 0.008), SWV_mean (p = 0.02), tumor size (p = 0.009), and Ki-67 (p = 0.03) on univariate regression analysis. There was no association between LVI status and other prognostic factors including histologic type, histologic grade, molecular subtypes, ER, PR and HER-2 (p > 0.1) (Table 4). On multivariate analysis, only EH value and Ki-67 were associated with LVI status (p = 0.003 and 0.025, respectively) (Table 5). For tumors > 2 cm, there was no association between LVI status and any clinicopathological variables on univariate regression analysis (p > 0.1) (Table 4).

Table 4.

Results from univariate regression analysis of clinicopathological variables for lymphovascular invasion in two subgroups divided by 2 cm

Clinicopathological variables	≤2 cm cancers		>2 cm cancers
	β coefficienta	p- value	β coefficienta	p- value
Age	−0.013 ± 0.01	0.37	0.02 ± 0.02	0.28
Tumor size	0.10 ± 0.04	0.009	0.03 ± 0.05	0.61
EH value	0.27 ± 0.09	0.003	−0.01 ± 0.12	0.91
SWVmax	0.19 ± 0.07	0.008	0.01 ± 0.10	0.91
SWVmean	0.25 ± 0.11	0.02	−0.001 ± 0.15	1.00
Histologic type
IDC	0		0
ILC	−20.46 ± 13397.66	1.00	−20.47 ± 40192.97	1.00
Mixed	0.34 ± 0.66	0.61	0.18 ± 0.66	0.79
Other	−20.46 ± 16408.71	1.00	0.74 ± 1.02	0.47
Histologic grade
I-II	0		0
III	0.22 ± 0.28	0.43	−0.13 ± 0.40	0.75
ER	0.003 ± 0.004	0.47	0.007 ± 0.005	0.16
PR	−0.002 ± 0.004	0.54	0.006 ± 0.005	0.20
HER-2	0.30 ± 0.30	0.32	−0.04 ± 0.41	0.93
Ki-67	0.01 ± 0.007	0.03	−0.007 ± 0.009	0.43
Molecular subtypes
Luminal A	0		0
Luminal B (HER2-)	0.55 ± 0.36	0.13	0.66 ± 0.64	0.30
Luminal B (HER2+)	0.56 ± 0.41	0.18	0.41 ± 0.71	0.57
HER2 enriched	0.54 ± 0.55	0.33	0.29 ± 0.83	0.73
Basal-like	−0.23 ± 0.57	0.69	−0.98 ± 1.21	0.42

EH, elastic heterogeneity; SWV, shear wave velocity.

^aData are expressed as estimates ± standard error.

Table 5.

Results from multivariate regression analysis of clinicopathological variables and lymphovascular invasion in cancers no more than 2 cm

Clinicopathological variables	β coefficient	p-value
Tumor size	0.10 ± 0.04	0.29
EH value	0.28 ± 0.09	0.003
SWVmax	0.19 ± 0.07	0.89
SWVmean	0.25 ± 0.11	0.88
Ki-67	0.02 ± 0.007	0.025

EH, elastic heterogeneity; SWV, shear wave velocity.

Since EH significantly correlated with LVI status especially for breast cancers ≤ 2 cm, diagnostic performance was evaluated for EH in distinguishing breast cancers ≤ 2 cm with LVI and those without. Using the most effective threshold of 1.36 m/s, the AUC, sensitivity, specificity, PPV and NPV were 0.61 (95 % CI: 0.55–0.67), 93.24% (95 % CI: 84.27–97.49) and 23.12% (95 % CI: 17.21–30.26), 34.16% (95 % CI: 27.74–41.19) and 88.89% (95 % CI: 75.15–95.84), respectively (Figures 3 and 4). Among 45 cancers ≤ 2 cm with negative EH, no LVI involvement was found in 89% of them (40/45), all of which were without secondary LN involvement (Figure 5). Besides EH value, Ki-67 was also positively correlated with LVI in tumors ≤ 2 cm at multivariate regression analysis, so Ki-67 was used to predict LVI. Using the most effective threshold of 8%, the AUC was 0.588. The sensitivity, specificity, PPV and NPV were 97.44% (95 % CI: 91.0–99.7), 17.51% (95 % CI: 12.38–24.10), 34.23% (95 % CI: 28.10–40.93) and 93.94% (95 % CI: 78.38–98.94) respectively (Figure 3).

Figure 3.

The ROC curve for EH and Ki-67 in differentiating tumors with LVI or not in breast cancers ≤ 2 cm. (a). The AUC for EH was 0.61. (b). The AUC for Ki-67 was 0.588. AUC, area under curve; EH, elastic heterogeneity; LVI, lymphovascular invasion; ROC, receiver operating characteristic.

Figure 4.

B-mode and VTIQ elastography maps in a 43-year-old breast cancer patient with LVI diagnosed by surgery. (a) B-mode map showed a BI-RADS 4C breast lesion with a maximum diameter of 14 mm. (b) The lesion (Solid white arrows) coded green on the quality map. (c) The SWV_max1, SWV_max2, SWV_max3, SWV_min1, SWV_min2 and SWV_min3 were measured as 9.22 m/s, 6.78 m/s, 7.19 m/s, 3.52 m/s, 2.62 m/s and 3.75 m/s respectively and the EH value was calculated as 4.43 m/s, which was higher than the cut-off value (1.36 m/s) for cT1-stage breast cancer. (d) The lesion was LVI-positive which was confirmed by pathological examination (hematoxylin-eosin stain; original magnification,×200) (Red arrows). EH, elastic heterogeneity; LVI, lymphovascular invasion; SWV, shear wave velocity; VTIQ, virtual Touch tissue imaging quantification.

Figure 5.

B-mode and VTIQ elastography maps in a 42-year-old breast cancer patient without LVI. (a) B-mode map showed a BI-RADS 4B breast lesion with a maximum diameter of 14 mm. (b) The lesion (Solid white arrows) coded green on the quality map. (c) The SWV_max1, SWV_max2, SWV_max3, SWV_min1, SWV_min2 and SWV_min3 were measured as 2.76 m/s, 2.69 m/s, 2.65 m/s, 2.52 m/s, 2.61 m/s and 2.65 m/s respectively and the EH value was calculated as 0.11 m/s, which was lower than the cut-off value (1.36 m/s) for cT1-stage breast cancer. (d) Pathological results showed the absence of LVI within and around the tumor. (hematoxylin-eosin stain; original magnification,×200). EH, elastic heterogeneity; LVI, lymphovascular invasion; SWV, shear wave velocity; VTIQ, virtual Touch tissue imaging quantification.

Discussion

The results of the study showed that lesions of breast cancer with poor prognostic features had higher SWVs than those with good prognostic features. In our study, higher SWV_max and SWV_mean were associated with larger tumor size (>2 cm), LVI, and axillary LNM (p ≤ 0.05 for all)—results that were consistent with previous studies.^13,15 Tumors > 2 cm, cancers with LVI, and cancers with axillary LNM had higher EH values than tumors ≤ 2 cm (p < 0.001), cancers without LVI (p = 0.017), and cancers without axillary LNM (p = 0.001) respectively. No statistical significance among EH values of molecular subtypes were observed (p = 0.09), which was different from SWV_mean (p = 0.005) and SWV_max (p = 0.014). Basal-like cancers had the lowest SWV_mean and SWV_max compared with other subtypes cancers.

In contrast to previous studies,^15,24 no association was observed between SWV_mean of histological Grade I-II and SWV_mean of Grade III in our study (p = 1.0). The reason may be that the number of Grade I cases was too small, and Grade I (n = 4) and II (n = 183) cases were combined for comparison with Grade III cases (n = 163). All breast cancers with pathological grade were invasive ductal carcinoma in our study. In the study performed by Evans, et al,¹⁵ the mean stiffness for Grade 2 cancers (mean: 143 ± 55 kPa) was similar to that of Grade 3 cancers (mean: 147 ± 58 kPa), and in our research, the SWV_mean of Grade 1 and 2 cancers group (mean: 4.48 ± 1.30 m/s) was similar to that of Grade 3 cancers (mean: 4.48 ± 1.35 m/s).

In our study, a correlation between SWV_mean and LVI were shown on univariate regression analysis (p = 0.05 for all cases; p = 0.02 for tumors ≤ 2 cm) while there was no association between the SWV_mean and LVI on multivariate regression analysis (p = 0.99 for all cases; p = 0.88 for tumors ≤ 2 cm), which was consistent with previous studies.^14,15 Among all SWV parameters, multivariate regression analysis identified only EH value as independent predictor of LVI in our study (p = 0.017 for all cases; p = 0.003 for tumors ≤ 2 cm). The result indicated that compared to SWV_mean and SWV_max, the EH value was more representative for the heterogeneity of breast cancer in terms of stiffness, and thereby could predict LVI status of lesions by identifying their aggressiveness.

On the SWE map, the highest stiffness areas of invasive breast cancer often appeared at the periphery of the tumor.²¹ In pathology, high stiffness areas often corresponded to tumor-associated collagen matrix and tumor–matrix interface.¹⁷ LVI usually occurred at the periphery of the tumor and tumor-matrix interface as well.^2,25 It was generally believed that tumor-associated collagen matrix usually acted as the “soil” seeding tumor occurrence and invasion, and the more collagen matrix there was, the more likely tumor cells could infiltrate.^14,26 On the other hand, the SWV values within the tumor tended to be low, appearing as soft regions on the SWE map. In our study, the EH value was determined by the difference between the average of the three maximum SWV values and the average of the three minimum SWV values. When the lesion had more tumor-associated collagen matrix and the tumor–matrix interface, there would be more stiff regions on the SWE map, and the EH value would be higher. Although not directly proven, it was very likely that LVI was more prone to occur in lesions with higher EH values.

In this study, although univariate analysis showed tumor size was significantly associated with LVI in tumors ≤ 2 cm (p = 0.009), there was no association between tumor size and LVI, suggesting that tumor size was not a pre-operative predictor for LVI. The positive rate of LVI in tumors ≤ 2 cm in this study was 31%, which was similar with Tong’s study (29%), while it was significantly lower in tumors ＞2 cm in this study (33%) than in Tong’s study (44%), since the tumor size range in Tong’s study was 0.5–8 cm compared to 0.5–3.5 cm in this study.²⁷ On the other hand, although some previous studies showed that positive rate of LVI in tumors ＞2 cm was significantly higher than in tumors ≤ 2, multivariate analysis was performed in few of these studies.^11,28 In this study, multivariate analysis showed no correlation was found between tumor size and LVI in all of the tumors and tumors ≤ 2 cm, which was consistent with Tong’s study. Furthermore, a recent study showed no significant difference was found in tumor size between breast cancers with LVI and those without, which indicated that further study was needed to analyze the correlation between LVI and tumor size with a large patient population.²⁹

Since EH was an independent predictor for LVI status especially for breast cancers ≤ 2 cm, diagnostic performance was evaluated for EH in distinguishing breast cancers ≤ 2 cm with LVI and those without. This study showed that in patients with cancers ≤ 2 cm (n = 255), although the AUC is low (0.61), the sensitivity and NPV could reach to 93.24 and 88.89%, respectively, which is similar with previous study.³⁰ It suggests that for T1-stage patients, LVI is likely to be negative when EH is negative. In this study, among the 45 tumors ≤ 2 cm with EH lower than the cut-off value of 1.36 m/s, no LVI involvement was found in 40 tumors, all of which were with no secondary lymph node involvement. For these patients, unnecessary aggressive ALND should be avoided. The sensitivity was high while specificity was low, this may be because EH was used to predict LVI indirectly by reflecting the spatial heterogeneity of elasticity in breast cancers. And EH value was also affected by other factors such as the proportion of tumor necrosis and the degree of fibrosis within the tumor. Necrotic areas corresponded to lower elastic values, while fibrotic areas corresponded to higher elastic values on elastography.³¹ EH was determined by gradient of stiffness. The EH value would be low when breast cancer had high proportion of tumor necrosis and low degree of fibrosis, and vice versa.

There are some limitations. First, this study only included mass forming breast cancer, which may have caused selective bias in the collection of cases. Second, only the largest plane of the lesion was used to calculate the SWV values. The results might have been more comprehensive if the 3D images of the breast lesions were used for analysis. Third, this study only included lesions with a maximum diameter ≤3.5 cm due to the limited width of the transducer.

Conclusions

Quantitative measurement of tumor elastic heterogeneity on SWE has the potential to be served as an imaging biomarker to predict lymphovascular invasion in breast cancer especially for tumors smaller than 2 cm.