Fifty-two women with 53 solid breast masses comprised the study group. The age range was 18 to 84 years, and the mean age was 53 years. Forty-seven lesions were symptomatic while six lesions were detected at mammographic screening. The mean ultrasound size of the symptomatic lesions was 21 mm and the screening detected lesions of 12 mm. The BI-RADS scores were: BI-RADS 2, two lesions (4%); BI-RADS 3, 20 lesions (38%); BI-RADS 4, 15 lesions (28%); and BI-RADS 5, 16 lesions (30%). Histology revealed 23 benign lesions and 30 cancers (28 invasive cancers, one DCIS and one lobular carcinoma in situ) in the study group.
The correlation between measurements by two independent observers of the mean stiffness on each pair of elastography images acquired by a single operator is shown in Figure 1. The intraclass correlation coefficient is 0.99, indicating that deciding which area on an image has the highest mean stiffness and measuring this area was highly reproducible. Three of 53 patients (6%) fell into different benign/malignant classifications when comparing the two observers. These three cases had mean stiffness values close to the 50 kPa cut-off value. The average difference in mean stiffness between the observers in these three discordant cases was 6 kPa. Intraclass correlation coefficients between two observers for maximum stiffness and SD, averaged from two images, were 0.98 and 0.93, respectively.
The correlation between measurements of the mean stiffness on pairs of elastography images acquired and read by two independent operators is shown in Figure 2. The intraclass correlation coefficient was 0.80, indicating moderate agreement. In two (13%) of the 15 patients the stiffness measurements taken from the images taken by two operators fell into nonconcordant benign/malignant classifications.
The elastography parameters of the 20 women with normal clinical and ultrasound findings are presented in Table 1. This indicates that normal tissue has low mean and maximum stiffness. The low SD values indicate low levels of tissue heterogeneity. Stiffness measurements of fatty and parenchymal tissue were not obviously different.
Fibroadenomas show low mean stiffness (average 28 kPa, range 18 to 44 kPa), maximum stiffness and SD values (Figures 3 and 4). The four benign lesions with mean stiffness above the 50 kPa threshold were two radials scars (both BI-RADS 4), one fat necrosis (BI-RADS 4) and one cellular fibroepithelial lesion (BI-RADS 3) removed to exclude a phyllodes tumour. Three of the four benign lesions with high elasticity values had core biopsy results showing lesions of uncertain malignant potential. The two cases of radial scar and the fat necrosis case were also classified as either BI-RADS 4 or BI-RADS 5 on greyscale imaging.
Ductal carcinoma in situ
The one DCIS lesion had a mean stiffness in the lower part of the malignant range (76 kPa). This lesion was a mass classified as BI-RADS 3.
Lobular carcinoma in situ
The single lobular carcinoma in situ case was in a mammographic distortion with malignant BI-RADS and elastographic features (mean stiffness, 82 kPa).
Twenty-seven (96%) of the 28 invasive cancers had mean elasticity values above the 50 kPa threshold (Figures 5 and 6). The average mean value was 140 kPa (range 29 to 293 kPa). Invasive cancers with an ultrasound size <15 mm had an average mean elasticity of 109 kPa, compared with an average value for lesions ≥ 15 mm of 167 kPa. The lesion with a mean stiffness below the 50 kPa threshold was a 12 mm grade 2 invasive ductal carcinoma of no special type, which was classified as BI-RADS 3. The lesion was sited very high in the breast and abutted both the skin and the pectoral muscle with very little surrounding breast tissue. The malignant lesions misclassified by greyscale ultrasound were a single DCIS case and three cases of invasive ductal carcinoma.
Shear wave elastography performance for maximum elasticity was sensitivity 97%, specificity 78%, PPV 85%, NPV 95% and accuracy 89%. For the SD, the performance was sensitivity 83%, specificity 96%, PPV 96%, NPV 95% and accuracy 89%. The number of true positive results, false positive results, true negative results and false negative results for each parameter are shown in Table 2.
Strong associations were demonstrated between greyscale BI-RADS findings, shear wave elastography findings (mean elasticity, maximum elasticity and SD) and histology (P < 0.0001 for all: Fisher's exact test). For mean elasticity versus greyscale BI-RADS, the performance results were sensitivity 97% versus 87%, specificity 83% versus 78%, PPV 88% versus 84%, NPV 95% versus 82% and accuracy 91% versus 83% (Table 3). The performance of shear wave elastography outperformed that of greyscale BI-RADS in all standard aspects of performance, including accuracy, PPV, NPV, area under the curve on a ROC analysis (Figure 7), and Youden's Index; however, this difference did not reach statistical significance (P = 0.129, Mann-Whitney U test).