Ductal carcinoma in situ (DCIS) is a precursor of breast cancer, and accounts for around 20% of the newly diagnosed breast cancers in United States. Recent studies have revealed that not all DCIS will eventually evolve into invasive breast cancer, so there is a critical need to stratify risk in patients with DCIS to avoid possible over-treatment of DCIS. DCE-MRI has better performance in DCIS detection and extent determination (1). DWI is a non-invasive contrast-free MR technique and shows potentials to complete DCE-MRI for DCIS risk stratification. The main purpose of this study is to explore whether quantitative apparent diffusion coefficient (ADC) of pure DCIS at 3 Tesla is able to distinguish DCIS pathological grades so as to aid DCIS risk stratification. 3T DWI data of 30 surgical pathology confirmed pure DCIS in 29 consecutive female patients (age: 51±11 years) were included. 3T axial breast DWI was conducted (Trio, Siemens, Erlangen, Germany) prior to DCE-MRI using a fat-suppressed single-shot EPI sequence (TE/TR=102/5800ms, voxel=1.8x1.8x6mm³, b=0, 1000 s/mm², averages=4) with a 4-channel breast coil. ROIs of DCIS lesions were carefully drawn on either b1000 image or ADC map based on better lesion visualization by using DCE-MRI image as reference, and mean ROI ADC was recorded. The mean ADC values of DCIS lesions associated with different (1: low; 2: intermediate; 3: high) grades (Van Nuys prognostic index scoring index) were compared using Kruskal-Wallis test and Mann-Whitney U test with a significance level of 0.05. The patient and lesion characteristics are given in Table 1. The mean ADC of DCIS lesions was 1.26±0.19 x10^-3 mm²/s, 1.51±0.29 x10^-3 mm²/s and 1.13±0.26 x10^-3 mm²/s for low (n=5), intermediate (n=9) and high grade (n=16) respectively. Kruskal-Wallis test and the box plot (Fig. 1) show that DCIS lesions with different grades had significantly different mean ADC values (p=0.0105). Mann-Whitney U test show that the mean ADC of high-grade DCIS was only significantly lower than that of intermediate-grade (p=0.0042) but not low-grade (p=0.2006). In addition, the mean ADC of high-grade DCIS was also significantly lower (p=0.0057, Fig. 2) than the non-high grade DCIS (1.42±0.28 x10^-3 mm²/s, combined low and intermediate grades, n=14). In contrast, the mean ADC of low-grade DCIS was not significantly lower than the non-low grade DCIS (1.27±0.33 x10^-3 mm²/s, combined low and intermediate grades, n=25).There are still sparse studies to investigate the correlation between DWI biomarker and pathology of DCIS (2, 3). Iima et al (2) reported that the mean ADCs of non-low (high and intermediate) grade DCIS were significantly lower than those of low grade DCIS in 22 patients. Rahbar et al (3) reported that high nuclear grade DCIS could be distinguished from non-high nuclear grade ones by mean maximum lesion size on DCE-MRI and lower mean contrast-to-noise ratio (CNR) at b600 DWI image in 55 pure DCIS. Both studies were acquired at 1.5T. 3T DWI supposes to better characterize DCIS due to its high SNR compared to 1.5T. Consistent with those two previous studies, our results indicate the potential roles of DWI for DCIS grade differentiation and risk stratification. However, on the other hand, our results are different in some aspects. Compared to Iima’s results, we found that the mean ADC was able to distinguish non-high grade DCIS from high grade ones, while low and intermediate grade DCIS could not be distinguished from each other. In contrast to Rahbar’s study, our DWI was acquired prior to DCE-MRI and different b-values were used (1000 v.s. 600 s/mm²). We correlated DCIS ADC to comprehensive pathological grade rather than simply the nuclear grade of DCIS. The mean ADC of low-grade DCIS was smaller than that of intermediate-grade DCIS in our study, which might attribute to the small number of low-grade DCIS. This study is limited in the relatively small sample size of DCIS. The pathological features that account for the ADC difference between different grades should be further delineated in future studies.