Researchers and clinicians interested in the biophysical basis of diffusion contrast in breast tissue and optimization of DWI for cancer detection.Conventional breast MRI using T₂-weighted and dynamic contrast for cancer detection has high sensitivity but variable specificity. Addition of DWI may increase specificity^1,2. Although diffusivity (ADC) changes in tissue are commonly attributed to ‘cellularity’ variations, direct evidence from breast tissue studies is limited. Reported correlation between ADC and cellularity is inconsistent (r= -0.802 to 0.048) ^3,4. A recent diffusion microimaging study of fixed breast tissue ⁵ showed distinct diffusivity differences between gland components, similar to prostate tissue ⁶. In the prostate, ADC correlates more strongly with partial volume of gland components than with cellularity ⁷. In this study we investigate the correlation between mean diffusivity and cellularity in the gland-lobular component of breast tissue.Formalin fixed tissue specimens from eight patients were sampled with a 3-mm core punch, immersed in 0.2% v/v Magnevist, and imaged on a 16.4T Bruker system (5 × 12 mm birdcage RF coil, Micro5 gradient set: 5 G/cm/A, 22^oC). Imaging was performed using a 3D spin echo DTI sequence at 40µm isotropic resolution. Number of averages = 4, δ/Δ = 2/12 ms, TE/TR = 30/400 ms, b = 800 s/mm2 with six directions and a single ‘b = 0’ reference measurement. ROIs were drawn in mean diffusivity (MD) images that showed distinct anatomical structures closely matched to the corresponding histological sections (Fig. 1). Corresponding ROIs (n = 64) in histology images were analyzed using Image Pro Premier to measure the cellularity metrics nuclear area and nuclear count. Kruskal-Wallis and post hoc Mann–Whitney testing was used to assess differences in MD among different types of tissue. Spearman’s correlation was used to assess the relation between MD and cellularity metrics.Fig. 2 summarizes measurements of cellularity metrics and MD in the normal gland lobules (comprising epithelium and intralobular fibrous stroma) and cancer tissues. A Kruskal-Wallis test with post hoc Mann-Whitney (KW-MW) analysis showed significant differences (p<0.05) in MD between all tissue types. These results are in good agreement with a previous microimaging study of fixed breast tissue ⁵. The KW-MW tests showed no significant differences for nuclear area, but significant differences for nuclear count for three of the six pairs (Fig. 3). Fig. 4 shows the correlation of cellularity metrics with measured MD after pooling all measurements. A Spearman test showed moderate inverse correlation between MD and nuclear count, but weak (ρ= -0.26, p<0.05) correlation with nuclear area. In contrast, Onishi et al. ⁴ reported moderate (ρ= -0.54, p<0.022) correlation between nuclear area and ADC for mucinous breast carcinoma at a voxel size of 12 mm³. Our results are not directly comparable with clinical imaging studies due to different tissue and imaging conditions. In particular, typical clinical imaging voxels are likely to include, besides the glandular lobule tissue we analyzed, significant volumes of interlobular stroma and fat which both have low cell density relative to the gland lobules. In breast MRI, the presence of multiple tissue types having very different cell densities (and MR properties in the case of fat) confounds the assessment of a simple relationship between ‘cellularity’ and gross measures of water diffusion dynamics such as ADC and MD. Inconsistent use of fat suppression in DWI will further confuse any attempt to correlate tissue properties with parameters derived from analysis of signal attenuation. This study, which focuses specifically on the glandular component of breast tissue, confirms the presence of microscopic tissue types having distinct diffusion properties, but suggests only a moderate correlation between MD and cellularity measured as cell count.