Characterizing the Diffusion Properties of Prostate Tissue Using Paired MR Microscopy and Multidimensional Diffusion MRI

Adam Phipps; Nyoman D Kurniawan; Geoff Watson; Paul Sved; Samson Dowland; Eleftheria Panagiotaki; David Atkinson; Roger Bourne

doi:10.1002/mrm.70344

Characterizing the Diffusion Properties of Prostate Tissue Using Paired MR Microscopy and Multidimensional Diffusion MRI

Magn Reson Med. 2026 Jul;96(1):349-365. doi: 10.1002/mrm.70344. Epub 2026 Mar 24.

Authors

Adam Phipps¹, Nyoman D Kurniawan², Geoff Watson³, Paul Sved⁴, Samson Dowland⁴, Eleftheria Panagiotaki⁵, David Atkinson¹, Roger Bourne⁶

Affiliations

¹ Centre for Medical Imaging, Division of Medicine, University College London, London, UK.
² Australian Institute for Bioengineering and Nanotechnology, Centre for Advanced Imaging, The University of Queensland, Brisbane, Queensland, Australia.
³ New South Wales Health Pathology, Sydney, New South Wales, Australia.
⁴ School of Medical Sciences, Faculty of Medicine and Health, University of Sydney, Sydney, New South Wales, Australia.
⁵ Hawkes Institute, University College London, London, UK.
⁶ Sydney School of Health Sciences, Faculty of Medicine and Health, University of Sydney, Sydney, New South Wales, Australia.

Abstract

Purpose: To characterize the diffusion properties of prostate epithelium and stroma in benign tissue and cancer.

Methods: Paired MR microscopy (20 μm) and multidimensional diffusion MRI (dMRI) (160 μm, b = 1000-2000 s/mm², ∆ = 15-120 ms) were performed at 16.4 T on 17 fixed prostate tissue samples (14× benign, 2× Gleason 3 + 3, 1× Gleason 4 + 4). MR microscopy images were used to segment epithelial, stromal, and luminal components in each sample. For each dMRI sequence, aggregate epithelial and stromal signal contributions in benign tissue were estimated using a linear epithelium-stroma-lumen signal model. Four diffusion signal models were fit to these aggregate signals. Quality-of-fit was assessed using the small-sample corrected Akaike Information Criterion (AICc). Voxel-wise model fitting was also performed to compare parameter estimates in benign tissue and cancer.

Results: Aggregate dMRI signal estimates for both epithelium and stroma were best described by the Ball + Sphere model (lowest AICc). A higher sphere fraction (0.278 vs. 0.175) and lower ball-compartment diffusivity (0.611 vs. 0.943 μm²/ms) were estimated for epithelium compared to stroma. The ADC model provided the worst fit in both cases (highest AICc). For Gleason 3 + 3 cancer, ADC and Ball + Sphere parameter estimates were consistent with the values found for benign epithelium and stroma; however, raised sphere fraction estimates were seen in Gleason 4 + 4 cancer.

Conclusion: Direction-averaged diffusion in fixed prostate epithelium and stroma is well described by the Ball + Sphere model. The diffusion properties of epithelium in Gleason 3 + 3 cancer and benign tissue appear to be similar; however, a marked increase to the volume fraction of restricted water was found for Gleason 4 + 4 epithelium.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Algorithms
Diffusion
Diffusion Magnetic Resonance Imaging* / methods
Epithelium / diagnostic imaging
Humans
Image Processing, Computer-Assisted / methods
Male
Microscopy* / methods
Prostate* / diagnostic imaging
Prostatic Neoplasms* / diagnostic imaging
Prostatic Neoplasms* / pathology

Grants and funding

EP/S021930/1/Engineering and Physical Sciences Research Council funded UCL Centre for Doctoral Training in Intelligent, Integrated Imaging in Healthcare (i4health)