Analyzing the Texture of Suspicious Lesions in the Female Breast with ADC-mapping in DWIBS and DWI
Jana Tesdorff1, Frederik Laun2, Stefan Delorme1, Wolfgang Lederer3, Heidi Daniel4, Heinz-Peter Schlemmer1, and Sebastian Bickelhaupt1

1Radiology, German Cancer Research Center, Heidelberg, Germany, 2Medical Physics, German Cancer Research Center, Heidelberg, Germany, 3Heidelberg, Germany, 4Mannheim, Germany


Diffusion-weighted imaging (DWI) can be helpful to differentiate benign and malignant lesions in the female breast. We compared the diagnostic performance of conventional DWI and DWIBS (DWI with background suppression) derived ADC maps with different definitions of the region of interest used to measure the ADC value (1.5T Philips). Texture analysis of suspicious breast lesions was performed utilizing ADC mapping in 59 lesions. Statistical analysis revealed the highest accuracy for lesion differentiation if using the mean ADC-value calculated of three small regions-of interest in the DWIBS derived ADC.


Breast cancer is the major cause of cancer related death in women worldwide. Magnetic resonance imaging (MRI) has the highest sensitivity in providing morphologic and functional information of breast tissue1. In clinical routine, diagnostic breast MRI can be complemented with diffusion-weighted imaging (DWI) to detect normal and abnormal tissue properties². A derived DWI sequence called DWIBS (diffusion-weighted imaging with background suppression) has been suggested to offer additive diagnostic capabilities to DWI. The aim of this evaluation is to compare the diagnostic performance of ADC maps derived from DWI and DWIBS for texture analysis of suspicious breast lesions with different definitions of the region of interest (ROI).


59 female patients (mean age 58.4y, SD ±6.2y) with suspicious breast lesions on screening X-ray mammograms (BIRADS 4/5) were enrolled in this retrospective, institutional and governmental review board approved study. Preliminary results of the study have been reported previously³. Prior to biopsy, breast MRI was performed on a 1.5T MR scanner (Ingenia, Philips) using a 7-channel breast coil acquiring non-enhanced morphological T1- and T2-weighted images, conventional epi DWI, DWIBS (diffusion-weighted imaging with dedicated background fat suppression) and contrast-enhanced MR images. Details for diffusion-weighted sequences are provided in Tab. 1. After transferring the images to a PACS workstation (OsiriX Imaging Software V.6.0, OsiriX Foundation, Geneva, Switzerland), apparent diffusion coefficient (ADC) maps were calculated applying a fit to SI=exp(-bi*D) with SI = images at b-values bi and D representing the apparent diffusion coefficient. This fit was a last-squares fit to log(SI) against bi. To quantify the ADC-value within the suspect lesion, the ROI function of the OsiriX software was used. A polygonal ROI was placed in every lesion in DWIBS and DWI at its biggest diameter. Three small ROIs, either round or oval depending on the lesion size and configuration, were inserted within each polygonal ROI particularly in lesion areas that visually presented the lowest signal intensity (Fig. 1). By means of SigmaPlot (Version 12.5, San Jose, USA) statistical analysis was performed. Receiver operating characteristic (ROC) curves and Friedman-Test for the analyses of variance were applied to compare the different methods of ADC-value quantification in benign and malignant lesions. A p-value <0.05 was assumed statistically significant.


Out of 59 evaluated lesions (mean diameter 11.8mm, SD ±7.2mm), 54% (n=32) were histopathologically proven to be malignant. Mean ADC-value in polygonal ROI was significantly lower (p<0.001) in DWI (880.8 10-6mm/s², 25-75% interquartile 745.0 10-6mm/s² - 1121.2 10 -6mm/s²) than in DWIBS (942.1 10-6mm/s², 25-75% interquartile 802.5 10-6mm/s² - 1164.0 10-6mm/s²). Areas under the curve (A) calculated by ROC curve analyses for the different methods of quantifying ADC-values in DWIBS and DWI are depicted in Fig. 2. The highest A was achieved using the mean ADC value calculated from three small ROIs within the lesion for both DWI (A=0.91) and DWIBS (A=0.96) with the DWIBS reaching a significantly higher A than the DWI derived ADC (p=0.03). The ADC measurement using a single polygonal ROI in DWIBS (A=0.93) and DWI (A=0.89) did not reach the A of the mean ADC value calculated of three small ROIs within the lesion and showed no statistically significant difference between the DWI and DWIBS sequence. This accounted as well for the assessment utilizing the small ROI with the lowest ADC-value in DWIBS (A=0.91) and DWI (A=0.90).


Despite the increasing use of diffusion weighted imaging derived ADC maps for lesion characterization in breast imaging, harmonized segmentation techniques to determine mean ADC value in a lesion have not been defined. We found that using the mean ADC value calculated out of three small ROIs placed within the lowest signal intensities in a lesion provided the highest diagnostic accuracy for lesion characterization in breast imaging using ADC maps. This ADC mapping technique was superior to both, the overall mean of a 2D region of interest (ROI) covering the largest diameter of a lesion and a solitary small ROI in its darkest spot. The methodological advantage accounted for both DWI and DWIBS, however DWIBS seemed to provide a slightly higher diagnostic performance than conventional DWI, which is in line with previous results suggesting the high diagnostic value of DWIBS in breast imaging4.


Characterization of suspicious breast lesions using DWI derived ADC maps was most accurately performed using the mean ADC value of three small ROIs placed in the areas that visually presented the lowest signal intensity. Both DWI and DWIBS could be used reliably to characterize suspicious breast lesions, but DWIBS seemed to provide a slightly higher overall accuracy.


No acknowledgement found.


1. Kuhl, Radiology 2007. 244:356-378.

2. Bammer, Eur J Radiol. 2003. 45:169-184.

3. Bickelhaupt et al., Radiology. 2015. Epub ahead of print.

4. Stadlbauer et al., Eur Radiol. 2009. 19:2349-56.


Table 1: Details for diffusion-weighted sequences used in the breast MRI protocol.

Figure 1: Case example of 59year old woman with histopathologically proven invasive breast cancer. A: T2-weighted, non-enhanced morphological sequence showing hypointense tumor in right breast (red arrow). Extract of ADC-maps of conventional DWI and DWIBS depicting localisation of polygonal and small ROI within the lesion in B and C, respectively.

Figure 2: ROC-curves of the used methods for quantification of the ADC-value.

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