Patients and lesions characteristics
A total of 221 patients with 230 breast lesions were enrolled in this prospective study. 17 lesions found in 14 patients were excluded from the analysis (13 patients did not undergo recommended follow-up and 1 patient developed adverse reaction (vomiting) shortly after the contrast administration). Thus, 213 lesions were included in the analysis of the contribution of quantitative and qualitative CEUS parameters in differential diagnosis of breast lesions. In total, 146/213 (68,5%) lesions were benign and 67/213 (31,5%) malignant.
Basic patient and lesion characteristics are summarized in Table 1. Patients with a malignant tumor were older (median 65 years vs. 54 years) and their lesions were larger (median 17 vs. 10 mm). Predominantly higher grade tumors occurred in the group of malignant tumors (almost 50 % of patients had grade 2 tumors and 37 % of lesions were grade 3).
Of 146 benign lesion, 74 (74/146; 51 %) were verified by a pathologist (66 cases after core-needle biopsy, 8 cases after surgical extirpation). The other lesions (72/146; 49 %) were evaluated as BI-RADS category 3 and therefore were not histopathologically verified; no changes were detected during a follow-up over at least two years. Of the benign lesions, the most frequent was fibrocystic breast disease (FCD), which accounted for a total of 32/74 (43%) of histologically verified lesions.
Altogether, 67 lesions of breast cancer were examined and histopathologically validated. The most common histopathological type (40/67; 60 %) was invasive carcinoma of no special type (NST), followed by invasive lobular carcinoma (13/67; 19 %).
Table 1. Basic patient and lesion characteristics.
Characteristics
|
Benign lesions
|
Malignant lesions
|
|
n=146
|
%
|
n=67
|
%
|
Age (years)
|
|
|
|
|
median
|
54
|
|
65
|
|
range
|
18-89
|
|
26-89
|
|
Lesion size (mm)
|
|
|
|
|
median
|
10
|
|
17
|
|
range
|
3-34
|
|
5-50
|
|
Side
|
|
|
|
|
left
|
89
|
61 %
|
27
|
40.3 %
|
right
|
57
|
39 %
|
39
|
58.2 %
|
left and right
|
0
|
0 %
|
1
|
1.5 %
|
Tumor Grade
|
|
|
|
|
1
|
-
|
-
|
9
|
13.4 %
|
2
|
-
|
-
|
33
|
49.3 %
|
3
|
-
|
-
|
25
|
37.3 %
|
Histology of benign lesions
|
n=74
|
|
-
|
-
|
fibrocystic disease
|
32
|
43 %
|
-
|
-
|
benign proliferative breast disease
|
12
|
16 %
|
-
|
-
|
fibroepithelial tumor
|
11
|
15 %
|
-
|
-
|
intraductal proliferative lesion
|
10
|
14 %
|
-
|
-
|
others
|
9
|
12 %
|
-
|
-
|
Histology of malignant lesions
|
-
|
-
|
n=67
|
|
invasive carcinoma NST
|
-
|
-
|
40
|
60 %
|
invasive lobular carcinoma
|
-
|
-
|
13
|
19 %
|
others
|
-
|
-
|
14
|
21 %
|
|
|
|
|
|
|
Abbreviations: NST, invasive carcinoma of no special type
CEUS parameters and type of tumor
The parameters of the quantitative and qualitative analysis of the CEUS of breast lesions are summarized in Table 2.
Table 2. Quantitative and qualitative parameters of breast CEUS according to the type of the tumor.
|
Benign lesions (n=146)
|
Malignant lesions (n=67)
|
p-value
|
Quantitative parameters
|
|
|
|
|
|
TTP (s)
|
|
|
|
|
<0.001
|
mean/ median/ range
|
29/ 25/ 4-152
|
|
20/ 18/ 4-70
|
|
|
WIS (dB/s)
|
|
|
|
|
<0.001
|
mean/ median/ range
|
0.13/ 0.04/ 0-1.64
|
|
0.24/ 0.14/ 0-1.63
|
|
|
PI (dB)
|
|
|
|
|
0.105
|
mean/ median/ range
|
2.8/ 2/ 0.1-10.5
|
|
2.9/ 2.8/ 0-8.6
|
|
|
AUC (dB.s)
|
|
|
|
|
0.711
|
mean/ median/ range
|
149/ 102/ 5-602
|
|
142/ 126/ 1-492
|
|
|
Qualitative parameters
|
n=146
|
%
|
n=67
|
%
|
|
Type of vascularization
|
|
|
|
|
0.117
|
peripheral
|
117
|
80 %
|
47
|
70 %
|
|
peripheral + central
|
29
|
20 %
|
20
|
30 %
|
|
Perfusion homogeneity
|
|
|
|
|
0.234
|
homogeneous
|
12
|
8 %
|
2
|
3 %
|
|
heterogeneous
|
134
|
92 %
|
65
|
97 %
|
|
Enhancement degree
|
|
|
|
|
<0.001
|
poor/absent
|
71
|
48.6 %
|
7
|
10.4 %
|
|
intermediate
|
47
|
32.2 %
|
23
|
34.3 %
|
|
rich
|
28
|
19.2 %
|
37
|
55.2 %
|
|
Abbreviations: CEUS, contrast-enhanced ultrasound; TI, time intensity curves; TTP, time to peak; WIS, wash in slope; PI, peak intensity; AUC, area under curve.
Statistically significantly lower TTP values (on average by 9 s) and higher WIS values (on average by 0.11 dB / s) were observed in malignant tumors compared to benign tumors. A statistically significant difference was also found in the enhancement degree when a rich vascularity was detected in 55.2 % of malignant lesions but only in 19.2 % of benign lesions. There was no statistically significant difference in the character of the blood supply in the surrounding tissue of benign and malignant breast lesions.
The results of univariate logistic regression models quantifying the relationship between types of lesions (malignant, benign) and individual CEUS parameters are summarized in Table 3.
A statistically significant relationship was found between the parameters TTP, WIS and enhancement degree. For every 10s increase in TTP the probability of malignancy decreased by 35 %. On the other hand, for every 1 dB/s increase of WIS the probability of malignancy increased 6 times. Furthermore, the higher the enhancement degree the greater the risk of malignancy. A lesion with intermediate enhancement degree is 5 times more likely to be malignant than a lesion with poor or absent perfusion. A richly vascularized lesion is up to 13 times more likely to be malignant than a poorly perfused lesion.
Significant variables from univariate analysis were further analyzed in multivariate logistic regression with 2 models combining one qualitative and one quantitative parameter (Table 3). The first model included the parameters enhancement degree and TTP, the second model included the enhancement degree and WIS. Neither of the quantitative parameters (TTP or WIS) increased the accuracy of the enhancement degree in distinguishing between malignant or benign type of a lesion.
Table 3. Univariate and two multivariate logistic regression models for diagnosis of malignant lesion.
|
Univariate association with malignancy
|
Multivariate association with malignancy Model 1
|
Multivariate association with malignancy Model 2
|
|
n=213
|
Crude OR (95 % CI)
|
p-value
|
n=213
|
adjusted OR (95 % CI)
|
p-value
|
n=213
|
adjusted OR (95 % CI)
|
p-value
|
Quantitative parameters
|
|
|
|
|
|
|
|
|
|
TTP (10 s)
|
|
0.65
(0.50-0.84)
|
0.001
|
213
|
0.78
(0.59-1.03)
|
0.076
|
-
|
-
|
-
|
WIS (1 dB/s)
|
|
6 (1.71-21)
|
0.005
|
-
|
-
|
-
|
213
|
1.6
(0.42-6.09)
|
0.489
|
PI (1 dB)
|
|
1.04
(0.91-1.19)
|
0.575
|
-
|
-
|
-
|
-
|
-
|
-
|
AUC (10 dB x s)
|
|
1.00
(0.97-1.02)
|
0.698
|
|
|
|
|
|
|
Qualitative parameters
|
|
|
|
|
|
|
|
|
|
Type of
vascularization
|
|
|
|
|
|
|
|
|
|
peripheral
|
164
|
1.00
|
ref.
|
-
|
-
|
-
|
-
|
-
|
-
|
peripheral +
central
|
49
|
1.72
(0.89-3.33)
|
0.110
|
-
|
-
|
-
|
-
|
-
|
-
|
Perfusion
homogeneity
|
|
|
|
|
|
|
|
|
|
homogeneous
|
14
|
1.00
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
heterogeneous
|
199
|
2.91
(0.63-13.39)
|
0.170
|
-
|
-
|
-
|
-
|
-
|
-
|
Enhancement
degree
|
|
|
|
|
|
|
|
|
|
poor or absent
|
78
|
1.00
|
-
|
78
|
1.00
|
-
|
78
|
1.00
|
-
|
intermediate
|
70
|
4.96
(1.97-12.49)
|
0.001
|
70
|
4.1
(1.6-10.47)
|
0.003
|
70
|
4.83
(1.91-12.18)
|
0.001
|
rich
|
65
|
13.40
(5.35-33.59)
|
<0.001
|
65
|
10.38
(4.04-26.72)
|
<0.001
|
65
|
12.00
(4.57-31.52)
|
<0.001
|
Abbreviations: OR, odds ratio; CI, confidence interval; TI, time intensity curves; TTP, time to peak; WIS, wash in slope; PI, peak intensity; AUC, area under curve.
All above mentioned logistic regression models were compared using ROC curves, the results are given in Table 4. Sufficiently discriminatory models (AUCROC > 60 %) used the parameters enhancement degree, WIS and TTP. The highest AUC was observed with the parameter enhancement degree (AUCROC = 74.7 %, 95% CI: 67.8–81.6). Higher sensitivity (89.6%) than specificity (48.6%) of this model indicates that it has a better ability to distinguish malignant tumors. The other two models with the parameters WIS (AUCROC = 69.8%, 95% CI: 62.1-7.5) and TTP (AUCROC = 67.8 %, 95% CI: 60.2–75.5) are also more accurate in distinguishing malignant lesions. In the multivariate logistic regression models, the selected quantitative parameters did not significantly improve prediction of malignancy. The combination of these parameters with the enhancement degree slightly increased the area under the ROC curve (in case of TTP by 3.5%, in case of WIS by 1.0%) and increased specificity (ability to recognize benign tumors), while sensitivity (the ability to recognize malignant tumor) decreased.
Table 4. Logistic regression models
|
ROC analysis of logistic regression models
|
Univariate logistic regression models
|
AUCROC (%)
|
95% CI (%)
|
p-value
|
Senzitivity (%)
|
Specificity (%)
|
TTP
|
67.8
|
60.2–75.5
|
<0.001
|
77.6
|
52.7
|
WIS
|
69.8
|
62.1–77.5
|
<0.001
|
74.6
|
66.4
|
PI
|
56.9
|
49.0–64.8
|
0.105
|
85.1
|
37.7
|
AUC
|
48.4
|
40.3–56.5
|
0.711
|
92.5
|
18.5
|
Type of vascularization
|
55.0
|
46.5–63.5
|
0.242
|
29.9
|
80.1
|
Perfusion homogeneity
|
52.6
|
44.4–60.8
|
0.540
|
97.0
|
8.2
|
Enhancement degree
|
74.7
|
67.8–81.6
|
<0.001
|
89.6
|
48.6
|
Multivariate logistic regression models
|
AUCROC (%)
|
95% CI (%)
|
p-value
|
Senzitivity (%)
|
Specificity (%)
|
Model 1 (TTP + enhancement degree)
|
78.2
|
71.7–84.8
|
<0.001
|
86.6
|
63.7
|
Model 2 (WIS + enhancement degree)
|
75.7
|
68.5–82.8
|
<0.001
|
79.1
|
70.5
|
Abbreviations: AUC, area under curve; CI, confidence interval; TI, time intensity curves; TTP, time to peak; WIS, wash in slope; PI, peak intensity.