Effect of myelin water exchange on DTI‐derived parameters in diffusion MRI: Elucidation of TE dependence
The diffusion‐weighted signal in brain tissue is often observed to decay non‐mono‐exponentially, and it is generally accepted that the diffusion process is of a multi‐compartment nature 6. To characterize the process, the bi‐exponential model is proposed to decompose the diffusion signal into fast/slow diffusion components corresponding to extra/intracellular compartments 9. However, in brain tissue, the volume fractions of intracellular and extracellular compartments are approximately 80% and 20% respectively, while the bi‐exponential model estimates that only 20% of the signal is from the slow intracellular diffusion component 11. To rectify this contradicting outcome, many groups have proposed models beyond the bi‐exponential approach. The CHARMED model accounts for diffusion in 3D, in which the intra‐axonal compartment is treated as cylinder‐shaped diffusion, while the extra‐axonal component is modeled with anisotropic Gaussian diffusion 13. The ActiveAx model expands CHARMED in the sense that the axon orientation should not be known a priori 14. The NOODI model simplified ActiveAx and further considered the fiber orientation dispersion 15. The above‐mentioned models assume no intercompartmental water exchange. However, the non‐exchange condition is often not fulfilled in the in vivo measurements, and this may lead to the underestimated volume fraction of the intracellular space 16. One of few models that includes water exchange effect between compartments is the Kärger model, in which the exchange rate is assumed to be related to membrane permeability 17.
As an alternative to using analytical models, Monte Carlo methods can also be used to provide the microstructure bio‐information by simulating random walks in arbitrary environments, when analytical solutions are not available because of the complexity of the system. Monte Carlo methods have been used to investigate the relation between diffusion NMR signal and biological environments in many previous studies. For example, Szafer et al. 2 investigated the effects of cellular volume fraction, extracellular, and intracellular diffusion on ADC. Budde and Frank 18 proposed a biophysical model of neurite beading to study the undulation of cell membrane in ischemia. Lin et al. 5 simulated the changes of DTI‐derived parameter due to different pathologies at cellular level after traumatic brain injury. Very recently, Harkins and Does 19 proposed a myelinated axon model and investigated the subtle influence of myelin water on diffusion‐weighted signal.
The role of myelin in diffusion models has been understudied because myelin water has a very short transverse relaxation time compared to the typical diffusion time, which makes the measured diffusion‐weighted signal blind to myelin water. However, the myelin water can still influence the diffusion signal in a subtler way, which deserves further exploration. By varying the echo time (TE), Qin et al.