Abstract
ABSTRACT
Time‐dependent diffusion MRI (dMRI) using single diffusion encoding (SDE) is sensitive to water dynamics in biological tissues, yet interpreting its signals requires careful consideration of underlying microstructure. While prior work has focused on restricted/hindered diffusion and membrane permeation, additional exchange mechanisms such as diffusion‐mediated exchange between dendritic shaft and spines in gray matter (GM) remain understudied. Here, we hypothesize that water diffusion within impermeable spiny dendrites can produce time‐dependent SDE signals indistinguishable from those arising from permeative exchange; and assess to what extent spine density impacts estimates of exchange time. Using Monte Carlo simulations and analytical solutions from the narrow escape problem, we quantify spine–shaft and shaft–spine exchange times, revealing characteristic times (1–50 ms) comparable to permeative exchange estimates in the cortex. We show that a modified two‐compartment Kärger model accurately captures the time‐dependent SDE signal along spiny dendrites but yields exchange estimates that reflect total spine volume fraction rather than specific spine morphology. Simulations reveal that unaccounted diffusion‐mediated exchange from dendritic spines substantially biases Neurite Exchange Imaging (NEXI) estimates, inflating the apparent exchange rate and the inferred membrane permeability in proportion to spine volume fraction. Further, we propose an extended three‐compartment Kärger model (and its coarse‐grained version) incorporating both diffusion‐mediated exchange between dendritic shaft and spines and permeative exchange with extracellular space. Critically, while the three‐compartment Kärger model and its coarse‐grained version capture both exchange mechanisms, they cannot uniquely disentangle membrane permeability from spine volume fraction. Finally, we highlight the importance of accounting for dendritic spines when inferring membrane permeability from diffusion MRI exchange rates, as the additional membrane area contributed by spines increases the effective dendritic surface‐to‐volume ratio by approximately 24%–60% for realistic spine densities and, if neglected, leads to a corresponding overestimation of the true membrane permeability. These findings underscore the necessity of considering dendritic spine contributions when interpreting time‐dependent SDE data and caution against attributing exchange effects solely to membrane permeability. Our study further highlights the need for advanced acquisition and modeling approaches to differentiate permeative and diffusion‐mediated geometric exchange in GM.