Dual-Parametric MR Imaging with Read-Out Segmented Diffusion-Weighted and High Temporal Resolution Dynamic Contrast-Enhanced Imaging Improves the Differentiation of Malignant and Benign Breast Lesions
Bin Wu1,2, Yanqiong Chen2, Hui Liu3, Xu Yan3, Caixia Fu4, Dan Wang1, Jian Mao2, Dominik Nickel5, Berthold Kiefer5, Yajia Gu2, and Weijun Peng2

1Radiology, Shanghai Proton and Heavy Iron Center, Fudan University Caner Center, Shanghai, China, People's Republic of, 2Radiology, Fudan University Shanghai Cancer Center, Shanghai, China, People's Republic of, 3NEA MR Collaboration, Siemens Ltd, Shanghai, China, People's Republic of, 4Siemens Shenzhen Magnetic Resonance Ltd, Shenzhen, China, People's Republic of, 5Siemens Healthcare GmbH, Erlangen, Germany, Forchheim, Germany


We investigated the clinical value of a dual-parameter classification method in differentiating benign and malignant breast lesions using readout-segmented diffusion-weighted imaging (RS-DWI) and quantitative dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), and found they correlated with histological results.

Target audience

Clinicians interested in multi-parametric analysis of breast lesions.


To investigate the clinical value of a dual-parameter classification method in differentiating benign and malignant breast lesions using readout-segmented diffusion-weighted imaging (RS-DWI) and quantitative dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), and study their correlation with histological results.


Eighty-three patients with breast masses confirmed by mammography or ultrasound were scanned with DCE-MRI (a prototype TWIST-Dixon VIBE sequence was used to achieve a temporal resolution of 5.3 s)1 and RS-DWI (RESOLVE, b = 50, 800 s/mm2)2 in a 3T MR scanner (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany). Seven patients were excluded because no obvious lesion was detected or because neoadjuvant chemotherapy was prescribed. The apparent diffusion coefficient (ADC) was calculated inline, and the Ktrans was calculated by using the commercially available software package Tissue 4D (Siemens Heathcare, Erlangen, Germany). The contour encompassed the entire lesion, and the adjacent normal gland was manually delineated on the last phase of DCE as Volume of Interest (VOI) by using the ITK-SNAP tool (www.itksnap.org). The ADC map was rigidly registered into DCE images using a prototype registration package based on ITK (www.itk.org). For each ADC and Ktrans 1D histogram, the following parameters were calculated for the entire tumor volume: the mean, median, low quantile, upper quantile, kurtosis and skewness. A 2D histogram (Ktrans-1/ADC) was also generated for the entire tumor volume using a prototype dual-parameter mapping package, and then the 2D kurtosis and 2D skewness were calculated from the normalized Ktrans-1/ADC map. Each parameter was correlated with a pathologic result, and its Receiver Operating Characteristic (ROC) was calculated.


All the patients underwent breast lumpectomy or radical resection after MR imaging, within a time interval of less than one week. Among the 76 clinically significant breast masses, 58 turned out to be malignant tumors including 53 invasive ductal carcinomas (IDC) and six ductal carcinoma in situ (DCIS), one invasive lobular carcinoma and one metaplastic carcinoma. Eighteen benign breast diseases were identified, including eight fibroadenomas, six adenoses, two intraductal papillomas, two benign phyllodes tumors and one sclerosing adenosis. As shown in Fig.1, for ADC analysis, the Area under the Curve (AUC) value of the median (0.790, 95%CI 0.683~0.873) was statistically higher than other parameters in ROC test (mean: 0.749, low quantile: 0.679, upper quantile: 0.597, kurtosis: 0.536, skewness: 0.770); for DCE MR imaging, AUC value of upper quantile (0.839, 95%CI: 0.739-0.912) was statistically higher than other parameters (mean: 0.794, median: 0.787, low quantile: 0.600, kurtosis: 0.661, skewness: 0.645); for dual-parametric Ktrans-1/ADC 2D histogram approach, the highest AUC of both 2D kurtosis (0.920, 95%CI: 0.837-0.969) and 2D skewness (0.919, 95%CI: 0.835-0.968) was achieved.

Discussion and Conclusion

The major limitation of DCE MR imaging evaluation in breast disease is that, in certain cases, benign lesions like fibroadenoma can also cause a local perfusion increase. Among all MR parameters that were evaluated as adjunct to DCE MR imaging, DWI emerged as the most robust and easy to use in clinical practice. However, certain malignant lesions were found with similar mean ADC as benign lesions but with higher perfusion, as shown in Fig. 2. The implementation of dual-parametric MR imaging in combination with DCE MR imaging and DWI optimizes the diagnostic accuracy in our study of breast tumors at 3T. The 2D kurtosis and skewness of Ktrans-1/ADC outperforms the single ADC or Ktrans analysis in differentiation of malignant and benign breast lesions with few overlapping in the same patient population. Further investigation on the clinical usage of dual-parameter analysis in a larger population base is a necessity, and it also might be useful in classifying pathological subtypes of breast cancer and monitoring the changes of neoadjuvant chemotherapy.


No acknowledgement found.


[1]. Mann RM, et al, Invest Radiol. 2014;49(9):579-85

[2]. Wolfgang Bogner, et al, Radiology, 2012; 263(1):64-76

[3]. Katia Pinker, et al, Radiology, 2015;276(2): 360-370


Fig 1: The ROC plot of 1D histogram analysis results for ADC (a) and Ktrans (b) single parameter. (c) The ROC plot of 2D histogram of Ktrans-1/ADC map for the entire tumor volume.

Fig 2: Ktrans-1/ADC 2D histogram map of benign (a,c) and malignant (b,d) lesions. The last phase DCE, realigned ADC and Ktrans map is in the left panel of each inlet. (a,b) gives the case for which Ktrans alone fails to differentiate between malignant and benign, but where ADC resolves the ambiguity. (c,d) shows the opposite case for which ADC fails to differentiate between malign and benign, but the Ktrans resolves.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)