Novikov, Fieremans, Jespersen, and Kiselev organize diffusion MRI (dMRI) biophysical modeling around coarse-graining over the diffusion length L(t)~βDt, defining three regimes: (i) short-time diffusion dominated by surface-to-volume ratio, (ii) time-dependent disorder-driven approach D(t)βDβ with a universal power-law exponent π=(p+d)/2 tied to structural universality class, and (iii) the long-time limit where each nonexchanging compartment is anisotropically Gaussian β the "Standard Model" (SM) of sticks + hindered Gaussian extra-neurite space convolved with a fiber ODF Novikov et al. 2018.
Two contributions stand out for a critical reader. First, the review's skeptical stance on its own field: it debunks the widely used empirical OGSEβPGSE diffusion-time conversion t=1/(4Ξ½) as "wrong," argues anomalous diffusion is empirically absent in brain (short-range structural disorder), and shows that axonal diameter mapping is dominated by extra-axonal (ln t)/t disorder effects β explaining the 3β15Γ histological overestimation of axon diameters in humans Novikov et al. 2018.
Second, its rigor on identifiability: the parameter count Np = Ns + l_max(l_max+3)/2 grows as lΒ² while cumulant information grows as lΒ³, yet degeneracy persists β at l_max=4 (DKI level) a continuous "trench" plus a discrete two-branch (bimodal) degeneracy make SM fitting underdetermined; only l_max=6 moments can lift both Novikov et al. 2018.
The review's framework is validated with Monte Carlo simulations, phantoms, and in vivo human/animal data (e.g., 45β600 ms time-dependent diffusion in human WM; human WM stick b-1/2 tail). Its emphasis on exchange and degeneracy directly motivated later clinical work: NEXI estimates intra/extra-neurite diffusivities and exchange time (~30β60 ms) in human cortex on 3T scanners, corroborating the review's caveat that nonexchange may fail in GM Jelescu et al. 2024. Orthogonal encodings (tensor-valued, planar) the review advocates remain actively optimized Szczepankiewicz et al. 2021, and distribution-based generalizations of the SM now embed it in maximum-entropy estimation frameworks MaxEnt-DTD 2026.
Verdict: Confidence high that this is field-defining (coarse-graining lens + honest degeneracy analysis); moderate confidence in the empirical universality-class claims outside WM, as the review itself notes only d=1 and d=2 phantom/MC validations. Falsification would require showing D(t) time-dependence not following the (p+d)/2 power law, or an observed non-Gaussian fixed point in brain tissue.
Simulation only: D(t)=Dβ+AΒ·t^-π with Dβ=0.72 Β΅mΒ²/ms (illustrative extra-axonal transverse tortuosity), A=0.9, π=1/2 (d=1, short-range disorder), Β±15% band reflecting unknown non-universal prefactor. Assumes nonexchanging stick+ens compartments and neglects exchange/permeability; not observed data.
Generated scientific data; not direct experimental measurements.Know what changed, what holds up, and what remains uncertain. Every Friday. No ads.