Ex vivo diffusion MRI characterizes tissue components in post-neoadjuvant rectal cancer specimens

Background: Despite advances in organ-preserving strategies for rectal cancer, accurate restaging after neoadjuvant therapy (NAT) remains challenging due to the limited sensitivity of conventional MRI in differentiating residual tumour from treatment-induced changes. This limitation highlights the urgent need to develop better imaging tools that can accurately analyze the complex structure of the treated rectal wall. Purpose: To study the diffusion properties of different rectal wall components, including healthy layers and pathological tissue, using high-resolution ex vivo diffusion MRI (dMRI) on whole total mesorectal excision (TME) samples obtained after NAT, and to evaluate how advanced diffusion metrics improve tissue analysis compared to standard T2-weighted imaging. Materials and Methods: Five post-NAT TME specimens were prospectively collected at a single center and fixed (36h formalin, 4h PBS). Then, specimens were mounted in Fomblin and scanned using a 9.4T Bruker BioSpec (22{degrees} ; 86 mm Tx/Rx). Diffusion MRI was acquired using a 2D multi-shell sequence (TR/TE 11,000/24 ms; 130 slices; 0.5 mm3 isotropic voxel; b = 1500 and 3000 s/mm; 15 directions) alongside multi-echo T2;-weighted imaging (TR 25,000 ms; 8 echoes; TE 10-80 ms; fat suppression). Diffusion and kurtosis parametric maps were generated by voxelwise linear least-squares fitting; T2 maps by monoexponential fitting (MATLAB). Specimens were sectioned at 5 mm, stained with H&E and dual staining (for fibrosis and smooth muscle), digitized, and co-registered with MRI using morphological landmarks. Regions-of-interest (ROIs) - mucosa, submucosa, muscle layers, tumour, and fibrous tissue - were compared using a linear mixed-effects model with FDR correction (RStudio v2025.09). Results: The muscularis propria exhibited the highest FA values of all tissue components, reflecting the ordered fiber architecture of its inner circular and outer longitudinal layers, which were visually separable on direction-encoded colour FA maps. Focal disruption of anisotropy at the tumour-muscle interface corresponded histologically to tumour invasion of the muscularis propria. Tumour regions showed the lowest mean diffusivity (MD), reflecting high cellularity and restricted diffusion, and MD was comparatively higher in the residual scar. Kurtosis metrics - particularly MK and AK - were elevated in tumour, reflecting greater microstructural heterogeneity and complexity. T2 mapping provided limited contrast across tissue types due to formalin fixation effects. Conclusion: Diffusion MRI metrics quantitatively discriminated rectal wall tissue components ex vivo with histological validation, beyond T2-weighted contrast. DTI and DKI metrics characterized tumour, fibrous tissue, and muscularis propria invasion, supporting their potential as microstructural imaging biomarkers for treatment response assessment.