Diagnostic value of contrast-enhanced ultrasound and shear-wave elastography for small breast nodules

Study design and patients

This retrospective study enrolled all patients with breast nodules who underwent conventional ultrasound, CEUS, and SWE at Gongli Hospital between November 2015 and December 2019. The Ethics Committee of Gongli Hospital approved this study (#[2020] Provisional Trial No. (003)). Given its retrospective design, the requirement for individual consent was waived by the committee. All patient details have been de-identified. The reporting of this study conforms to the STROBE guidelines as per the recommendation (von Elm et al., 2007).

The inclusion criteria encompassed patients who met the following criteria: (1) a breast nodule with a diameter of ≤2 cm with pathological outcomes based on the results of surgical or needle biopsy, (2) no prior treatment for a breast nodule before ultrasonography, and (3) a solid or predominantly solid breast nodule (with the cystic component constituting <25% of the total volume). The breast density of the patients included in this study comprises Type B, Type C, and Type D as determined by ultrasound characteristics. These types reflect a range of breast tissue compositions, including predominantly fibrous (Type B), heterogeneous density with both fibrous and glandular elements (Type C), and predominantly glandular tissue (Type D), which can influence the sensitivity of imaging techniques and the interpretation of findings.

Conversely, the exclusion criteria were (1) pregnancy or lactation and (2) low-quality breast nodule images that were unsuitable for analysis, including cases with blurriness on conventional ultrasound or inconclusive SWE measurements.

For clarity, patients with BI-RADS category 1 lesions, indicating no abnormalities, were excluded from the study. Additionally, patients with BI-RADS category 2 lesions were also excluded. The rationale for this exclusion is that category 2 nodules are typically benign and they often require no further investigation beyond regular follow-up. It is only when these nodules exhibit changes that might warrant a change in the BI-RADS classification that additional imaging or biopsy is considered. A visual representation of the patient selection process is detailed in Fig. 1.

Data collection and imaging examinations

Information on age, tumor size, location, shape, orientation, margins, echo patterns, posterior acoustic features, calcification, vascularity, and lymph node metastasis was retrieved from the patients’ medical records. The multimodality ultrasound examinations were all performed on the same day.

Conventional ultrasound

Operators recorded comprehensive details regarding nodular characteristics, encompassing location, size, shape, orientation, margin, echo pattern, posterior acoustic features, calcification, and vascularity. Subsequently, lesions were classified according to the BI-RADS classification system. In cases in which multiple nodules were present, the nodule with the highest BI-RADS classification was included in the study. If two nodules held the highest BI-RADS classification, both were included concurrently.

SWE

During the study period, a Siemens Acuson S3000 ultrasound diagnostic apparatus (Siemens Medical Solutions, Mountain View, CA, USA) equipped with SWE imaging software was employed for SWE. SWE was conducted when the gray-scale ultrasound indicated that the maximum diameter of the lesion and the image clarity was optimal. The SWE sampling frame size was adjusted to be at least twice the size of the nodules, and patients were instructed to briefly hold their breath after achieving image stability. The SWE speed mode was utilized to directly derive the shear-wave velocity (SWV) across the two-dimensional spatial distribution of the SWE imaging map. The SWV range was gradually fine-tuned (with a maximum of 10.0 m/s) when the interior regions of the nodules displayed red or yellow hues, whereas the surrounding areas appeared blue or green. Multiple regions of interest (typically 5–7) were strategically positioned within various areas within the nodules (upper, lower, middle, and periphery at the highest and lowest speeds). SWV measurements were taken within the effective measurement areas, and the average SWV (m/s) was subsequently determined for each nodule.

CEUS

During the study period, a Philips EPIQ 5 ultrasound diagnostic apparatus equipped with CEUS software (Philips Medical Systems, Bothell, WA, USA) was used. For CEUS, sections exhibiting robust blood flow, prominent blood vessels, or irregular shapes were selected. Areas featuring substantial calcification accompanied by broad sound shadows were intentionally avoided. The focal point was positioned behind the nodule using a mechanical index of 0.07. The contrast agent SonoVue (25 mg in five mL of 0.9% sodium chloride, Bracco SpA, Milan, Italy) was administered per the standard procedure. The nodule’s dynamic perfusion process was observed for at least 3 min. During CEUS, nine distinct variables were assessed (Luo et al., 2016): (1) enhancement intensity (low enhancement, equal enhancement, high enhancement); (2) order of enhancement (concentric, non-concentric); (3) change (difficult to discern, shrinking, unchanged, expanding); (4) enhancement uniformity (uniform, non-uniform); (5) enhancement defects (present or absent); (6) morphology after enhancement (regular, difficult to discern, irregular); (7) enhanced posterior boundary (clearly distinguishable, difficult to discern, unclear); (8) claw sign (present or absent); and (9) presence of nourishing blood vessels. Discrepancies in evaluations were resolved through discussion between two ultrasound physicians to reach a consensus.

Imaging analysis

All SWE image acquisitions and subsequent data analyses were performed by two radiologists with more than 3 and 10 years of expertise, respectively, in SWE and breast ultrasonography. CEUS images were evaluated by two radiologists with more than 5 and 10 years of proficiency, respectively, in CEUS and routine ultrasound examinations. Each ultrasound image was reviewed and confirmed by two ultrasound specialists not involved in the acquisition of the contrast images. The contrast characteristics of the lesions were categorized following consensus agreement. Prior to image analysis, these two ultrasound specialists were not privy to the patients’ clinical data, ensuring an unbiased evaluation process. This approach ensured consistent assessment of CEUS images by the two radiologists. Disagreements between the radiologists were resolved through joint re-evaluation of the imaging features. This collaborative process ensured that any discrepancies were resolved in a manner that maintained the integrity and reliability of the study findings.

To determine the most accurate classification for breast nodules, an integrated multimodal imaging approach was utilized. Initially, each nodule was given a preliminary classification based on the BI-RADS system. Subsequently, nodule vascularity was assessed using CEUS, and any nodule exhibiting at least two malignant features was classified as malignant. The classification was further refined using SWE, with nodules having SWVs of 3.7 m/s or higher being considered malignant. The final categorization was achieved by combining the results of the initial BI-RADS classification with the findings of CEUS and SWE. In cases of disagreement between CEUS and SWE, the BI-RADS classification was either upgraded, downgraded, or retained (Fig. 1). The resulting integrated methodology provided an enhanced diagnostic platform, amalgamating the insights of BI-RADS, CEUS, and SWE. The SWE classification cutoff 3.7 m/s is derived from our team’s clinical experience and a detailed analysis of patient data from the hospital. This value was established through receiver operating characteristic (ROC) curve analysis, which provided the highest accuracy in distinguishing between benign and malignant nodules.

Statistical analysis

The statistical analysis was performed using SPSS 22.0 (IBM, Armonk, NY, USA) and MedCalc 19.0.7 (MedCalc Software bvba, Ostend, Belgium). Continuous variables were reported as means ± standard deviations or ranges, and comparisons were made using the independent-samples t-test. Categorical data were presented as n (%) and analyzed using the chi-squared test or Fisher’s exact test. The agreement between the two radiologists for CEUS and SWE evaluations was determined using the interclass correlation coefficient (ICC), with values close to 1 indicating excellent reliability. To assess the diagnostic efficacy of classifying small breast nodules, ROC curves were employed for the four diagnostic methods (BI-RADS, CEUS, SWE, and the combined method). Areas under the curve (AUCs) were computed to determine the diagnostic performance of the four methods. For statistical evaluation, the Cochran Q-test and z-test were utilized. The optimal cutoffs were derived from the ROC analysis, with subsequent calculation of sensitivity (SEN), specificity (SPE), positive predictive value (PPV), negative predictive value (NPV), and accuracy (ACC). Statistical significance was defined as a two-sided P-value of less than 0.05.

Clinical data and pathological findings

In this study, of the 676 initially assessed patients, 302 met the inclusion criteria, and 305 nodules were analyzed because three patients had two nodules with the same highest BI-RADS classification. Among these, 113 nodules (37.0%) were malignant, and 192 (63.0%) were benign. The mean age of the patients was 49.2 ± 16.4 years. The benign nodules comprised various pathologies, most commonly fibroadenoma (43.2%), adenopathy (21.4%), and adenopathy with fibroadenoma (20.3%). The malignant nodules were mainly invasive ductal carcinoma (69.0%), followed by ductal carcinoma in situ and papillary carcinoma (9.7% each).

Ultrasound predictors of malignancy

Patient age, the breast nodule size, the breast nodule location, and the echo pattern did not exhibit significant differences between patients with benign and malignant nodules (all P > 0.05). However, notable distinctions were observed in terms of shape (P < 0.001), orientation (P < 0.001), margins (P < 0.001), posterior acoustic features (P = 0.001), calcification (P < 0.001), internal vascularity (P < .001), and lymph node metastasis (P < 0.001) between malignant and benign lesions (Table 1 and Fig. 2).

Diagnostic efficacy

The BI-RADS classifications of the breast nodules are outlined in Table 2. The determined cutoff for the BI-RADS classification system was category 4B. Correspondingly, this yielded SEN, SPE, PPV, NPV, ACC, and AUC of 82.3%, 74.5%, 65.5%, 87.7%, 77.4%, and 0.784, respectively.

Interrater reliability, as assessed using ICC, indicated excellent agreement between the two radiologists for both CEUS (ICC = 0.91, 95% confidence interval (CI) [0.89–0.93]) and SWE (ICC = 0.89, 95% CI [0.86–0.91]). The distinct characteristics indicative of potential malignancy in breast nodules identified by CEUS are presented in Table 3. CEUS demonstrated the capacity to identify malignant nodules when at least two of the nine suspicious malignant signs were concurrently present. In this scenario, the ensuing SEN, SPE, PPV, NPV, ACC, and AUC were 83.2%, 87.5%, 79.7%, 89.8%, 85.9%, and 0.853, respectively.

The diagnostic efficiency of SWE was determined using the mean SWV. Malignant nodules exhibited a mean SWV of 5.2 ± 1.6 m/s, which was significantly higher the value observed for benign nodules (3.1 ± 1.1 m/s, P < 0.001). Employing ROC curve analysis, the optimal cutoff for SWV was 3.7 m/s. Consequently, the resulting SEN, SPE, PPV, NPV, ACC, and AUC were calculated as 86.7%, 82.8%, 74.8%, 91.4%, 84.3%, and 0.848, respectively.

When CEUS or SWE alone was combined with the BI-RADS classification in diagnosing benign nodules, the BI-RADS classification remained unaltered. Nonetheless, the AUC for this combined diagnosis was 0.758, indicating lower performance compared to the individual BI-RADS classification, SWE, and CEUS. Conversely, the combination of CEUS and SWE led to a one-category increase and decrease in the BI-RADS classification for malignant and benign nodules, respectively. This combined diagnostic approach yielded SEN, SPE, PPV, NPV, ACC, and AUC of 88.5%, 87.0%, 80.0%, 92.8%, 87.5%, and 0.877, respectively.

Significant differences were noted among the four diagnostic methods, as indicated by the Cochran Q-test (Cochran’s Q = 19.573, P < 0.01). The AUCs for the diagnostic efficacy of BI-RADS, CEUS, SWE, and the combined method (CEUS + SWE + BI-RADS) were 0.784, 0.853, 0.848, and 0.877, respectively. In the statistical comparison of the combination diagnostic model with BI-RADS, CEUS, and SWE when utilized individually, P-values of <0.001, 0.012, and 0.017, respectively, were obtained. These results suggest that the combination diagnostic approach is significantly more effective than the individual methods (Tables 4–5 and Fig. 3). Implementing this combined diagnostic approach resulted in the reclassification of 55.8% (67/120) of the BI-RADS 4A nodules as BI-RADS 3. These reclassified nodules were subsequently confirmed as benign through pathological assessment, suggesting that unnecessary biopsy procedures could potentially be avoided for such nodules.

CONCLUSIONS

The integrated diagnosis using multiple ultrasound techniques (BI-RADS with conventional ultrasound, CEUS, and SWE) displayed an enhanced capability to differentiate benign and malignant breast nodules. CEUS and SWE might effectively serve as supplementary tools to enhance the clarity of the BI-RADS classification for smaller benign nodules, thereby refining the management of BI-RADS categorization for this subset of cases.