Research Article: Differentiation of malignant and benign breast lesions: Added value of the qualitative analysis of breast lesions on diffusion-weighted imaging (DWI) using readout-segmented echo-planar imaging at 3.0 T

Date Published: March 30, 2017

Publisher: Public Library of Science

Author(s): Yeong Yi An, Sung Hun Kim, Bong Joo Kang, Tone Frost Bathen.

http://doi.org/10.1371/journal.pone.0174681

Abstract

To determine the added value of qualitative analysis as an adjunct to quantitative analysis for the discrimination of benign and malignant lesions in patients with breast cancer using diffusion-weighted imaging (DWI) with readout-segmented echo-planar imaging (rs-EPI).

A total of 99 patients with 144 lesions were reviewed from our prospectively collected database. DWI data were obtained using rs-EPI acquired at 3.0 T. The diagnostic performances of DWI in the qualitative, quantitative, and combination analyses were compared with that of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Additionally, the effect of lesion size on the diagnostic performance of the DWI combination analysis was evaluated.

The strongest indicators of malignancy on DWI were a heterogeneous pattern (P = 0.005) and an apparent diffusion coefficient (ADC) value <1.0 × 10–3 mm2/sec (P = 0.002). The area under the curve (AUC) values for the qualitative analysis, quantitative analysis, and combination analysis on DWI were 0.732 (95% CI, 0.651–0.803), 0.780 (95% CI, 0.703–0.846), and 0.826 (95% CI, 0.754–0.885), respectively (P<0.0001). The AUC for the combination analysis on DWI was superior to that for DCE-MRI alone (0.651, P = 0.003) but inferior to that for DCE-MRI plus the ADC value (0.883, P = 0.03). For the DWI combination analysis, the sensitivity was significantly lower in the size ≤1 cm group than in the size >1 cm group (80% vs. 95.6%, P = 0.034).

Qualitative analysis of tumor morphology was diagnostically applicable on DWI using rs-EPI. This qualitative analysis adds value to quantitative analyses for lesion characterization in patients with breast cancer.

Partial Text

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is widely used in breast cancer diagnosis and staging. DCE-MRI has widely demonstrated diagnostic value in breast imaging [1]. It provides high-resolution morphological information regarding the contrast-enhanced characteristics of the lesions. Although DCE-MRI has a high sensitivity of 94–100%, the specificity is only 40–80% for the characterization of the breast [2–4]. In addition to this relatively low specificity, DCE-MRI is time consuming and expensive. Furthermore, it carries the risk of potential side effects from the contrast media. Therefore, non-contrast-enhanced imaging techniques have been actively investigated as alternatives or adjuncts to DCE-MRI to detect breast cancer [4–8].

The mean size of the malignant masses was 2.42 ± 1.78 cm, and the mean size of the benign masses was 1.03 ± 1.21 cm. The median sizes of the malignant and benign masses were 2.0 cm (range: 0.5–10.2 cm) and 0.75 cm (range: 0.3–6.8 cm), respectively. Three lesions were not visible on DWI: a 1.5-cm radial scar, a 0.6-cm intraductal papilloma, and a fibrocystic change of 0.4 cm. Of the 141 lesions visible on DWI, 119 were mass type lesions, and 22 were non-mass type lesions.

Our study demonstrated that qualitative DWI analysis based on morphological analysis was useful in predicting malignancy and has a potential to improve the diagnostic performance of DWI. Using multivariate analysis, the heterogeneous internal pattern of various morphological descriptor on DWI, such as that of a low ADC value (<1.0×10−3 mm2/sec), was the most significant independent predictor of malignancy. Additionally, the combined DWI analysis enabled improved diagnostics to predict breast cancer by increasing sensitivity without a loss of specificity in quantitative ADC analysis, although it was inferior to the combination of DCE-MRI and ADC. Currently, there have been few studies that evaluate the diagnostic usefulness of the morphological analysis of breast lesions on DWI [15–16]. Previously, Kang et al. investigated the diagnostic accuracy and usefulness of a high signal rim sign on DWI [15]. The sensitivities, specificities, and AUC values were 59.7%, 80.6%, and 0.701, respectively, for the rim sign and 82.3%, 63.9%, and 0.731, respectively, for the ADC value (cutoff ≤1.46×10−3 mm2/sec). Their results suggested that a high signal rim sign on DWI was a valuable morphological feature to improve specificity in DWI. However, they only focused on one morphological characteristic on DWI, the rim sign. We think that further evaluation of the diagnostic performance and positive predictive values of each morphological descriptor are necessary for the differentiation between benign and malignant lesions on DWI. Recently, Barentsz et al. examined the diagnostic value of qualitative analysis of DWI using the reduced field-of-view (rFOV) technique in 30 breast lesions [16]. In that study, the shape and BI-RADS classification of the lesions were considered in the qualitative analysis. The discriminative abilities based on ADC values were similar for ss-EPI and rFOV, with AUCs of 0.79 and 0.82, respectively. When the lesion shape was included in the analysis, the AUCs from the three readers ranged from 0.74 to 0.91 for rFOV and from 0.67 to 0.75 for ss-EPI. When the BI-RADS classification of the lesion was added to the interpretation, the AUCs for the three readers were 0.71–0.93 for rFOV, 0.61–0.76 for ss-EPI, and 0.87–0.91 for DCE-MRI. These results suggested that additional assessment of tumor morphology with rFOV contributed to the higher AUCs, which is consistent with our results. However, the rFOV technique has two major limitations: unilateral breast coverage and additional scan time as an adjunct to standard DWI. Therefore, we believe that the rs-EPI technique is advantageous over rFOV because it can cover the entire breast.   Source: http://doi.org/10.1371/journal.pone.0174681

 

0 0 vote
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments