Pancreatic ductal adenocarcinoma (PDAC) develops silently and is often detected in the late stages with only ~10-15% operable when diagnosed. The dense desmoplastic stroma in PDAC limits delivery of imaging probes for detection and drug delivery. Chronic pancreatitis has similar clinical behavior and imaging features as PDAC, making it harder to detect PDAC¹. Survival outcomes could be significantly improved by early detection. Diffusion MRI is a noninvasive imaging technique that is being increasingly used for breast and brain cancer detection and has been explored in the detection of PDAC, to improve the confidence level of detecting PDAC with MRI². The underlying changes in water diffusion in PDAC are not defined and their characterization would provide a strong rationale to incorporate this external label-free modality in PDAC detection. Diffusion tensor imaging (DTI) has been extensively used to examine tissue microstructure in neurological diseases³, such as white matter tracks, and we have recently observed heterogeneous DTI parameters in breast cancer. These DTI parameter patterns were seen to follow the underlying Collagen 1 (Col1) fiber distributions. We observed in regions with low Col1 fiber content that the apparent diffusion coefficient (ADC) and fractional anisotropy (FA) values decreased significantly as compared to regions containing dense Col1 fibers. Col1 fibers are a major structural component of the tumor extracellular matrix (ECM) and play an important role in cancer dissemination and molecular transport through the tumor interstitium. Col1 fibers are detected using second harmonic generation (SHG) microscopy ex vivo on tumor sections. Here, we have combined DTI and optical imaging to characterize diffusion parameters in a PDAC xenograft model to understand the relationship between Col1 fibers and diffusion in PDAC. We observed that regions containing fewer Col1 fibers were characterized by lower ADC and FA compared to dense fiber regions. Results here support the use of ADC and FA in detecting and characterizing PDAC.BxPC3 PDAC were implanted orthotopically in SCID mice as previously described⁴. Tumors were excised from euthanized mice at volumes of ~ 300 mm³ and fixed in 4% paraformaldehyde. A vertical 11.7 Tesla spectrometer was used to acquire three-dimensional DTI with two non-diffusion weighted images and eight diffusion-weighted images (b=1500 s/mm², resolution 60 x 60 x 60 μm³). Following DTI, the tumors were paraffin-embedded and sectioned at 5 μm thickness. The sections were then stained with hematoxylin and eosin (H&E). These H&E sections were used to acquire tiled scan SHG microscopy to detect the Col1 fibers distributions in 3D (incidence = 860 nm, emission = 430 nm) using a 25× lens on an Olympus FV1000 multiphoton microscope. ADC and FA maps were calculated from the DTI data. Multimodality co-registration was performed using affine transformation to co-register the SHG images to the diffusion images. We found that high Col1 fiber density correlated with increased ADC and FA in the tumors, compared to regions with fewer Col1 fibers (Figure 1). We observed that regions with aligned Col1 fibers had increased diffusion anisotropy (compare Figure 1D with E). The primary diffusion direction in high anisotropy regions mostly followed the underlying Col1 fiber distribution alignment. We simulated water diffusion directionality maps from the maximum intensity projected SHG data and the simulated results (in X-Y direction) displayed in Figure 1F closely matched the actual water diffusion directionality maps. Consistent with our previous observations in human breast cancer tissue, fewer Col1 fibers exhibited lower ADC and FA compared to dense fiber regions. These data suggest that delivery and transport of low molecular weight chemotherapy agents through the tumor ECM may be influenced by the Col1 fiber distribution. The results obtained here support further investigating the use of intrinsic ADC and FA to noninvasively detect PDAC. Diffusion-MRI can be performed on most clinical MRI scanners, providing ease of clinical translation. Our data create much-needed new possibilities for the detection of PDAC and for evaluating cancer architecture using diffusion MRI.