To understand the brain’s computational power, researchers aim to map its "wiring"—the billions of axons that transmit signals between neurons. While standard MRI can show the paths these wires take, it struggles to measure their thickness. Measuring axon diameter is critical because it dictates signal conduction velocity, yet most human axons are less than 1 micrometer ($\mu$m) wide. This creates a fundamental "resolution floor" where the physical limits of MRI hardware collide with the microscopic scale of biology.
The Sensitivity Gap
The core challenge of microstructural imaging is that diffusion MRI does not "see" axons directly; instead, it infers their size by tracking the random motion of water molecules. As water molecules bounce off the internal walls of an axon, their displacement is restricted. By measuring how much this restriction attenuates the MRI signal, researchers can estimate the axon's radius.
However, as Alexander demonstrated, there is a severe sensitivity gap. On standard clinical scanners, the diffusion signal does not change enough to distinguish between different small-scale radii. This creates a resolution floor—a size threshold below which all structures look identical to the scanner. Schneider et al. and Siow et al. note that for clinical hardware, this floor is approximately 2 to 5 $\mu$m. Since the majority of human axons are sub-micrometer, they remain effectively invisible to standard quantification.
The Fourth-Power Scaling Problem
Why is this floor so hard to break? The difficulty lies in the physics of signal attenuation. In the regime of restricted diffusion, the signal sensitivity scales with the fourth power of the radius ($r^4$). As Baraković et al. explain, this means that if an axon's radius drops by half, the signal change associated with that restriction drops by a factor of 16.
To detect such minute changes, the imaging hardware must provide incredibly strong magnetic field gradients ($G_{max}$). Dyrby et al. found that increasing gradient strength significantly improves the ability to differentiate axon sizes, but clinical scanners are often limited to 40–80 mT/m. Resolving the fine microstructures typical of human white matter requires "super-scanners" with ultra-strong gradients, such as the 300 mT/m systems described by Alexander et al..
Pushing the Limits of Waveforms
Beyond just raw power, the shape of the gradient pulse matters. The standard Pulsed-Gradient Spin-Echo (PGSE) sequence has an inherent window of sensitivity. To probe smaller structures, researchers use Oscillating Gradient Spin Echo (OGSE) sequences. By "packing" more diffusion weighting into shorter time scales, OGSE allows the MRI to probe microstructures that were previously obscured. Drobnjak et al. showed that while PGSE is efficient for idealized parallel fibers, OGSE offers superior sensitivity in realistic scenarios where fiber orientations are dispersed or unknown.
The Index vs. Absolute Measurement
Even with advanced hardware, a persistent problem remains: overestimation. Sepehrband et al. observed that MRI-based axon diameter estimates consistently exceed measurements from electron microscopy. This suggests that current models may be capturing a "diameter index"—a value that tracks relative biological trends—rather than an absolute physical dimension.
Because of the resolution floor, Schneider et al. argue that measurements of structures below the threshold should be treated as relative indices. To move toward absolute accuracy, newer methods are looking beyond pure diffusion. For example, Baraković et al. and Canales‐Rodríguez et al. have developed "diffusion-relaxation" models. By analyzing how water "relaxes" near cell surfaces rather than just how it moves, these models can potentially predict axon sizes below one micrometer, bypassing the $r^4$ sensitivity wall that limits traditional diffusion imaging.
Key Foundational Work
Axon radius measurements in vivo from diffusion MRI: a feasibility study (2007) — contributed 3 concept elements to this direction.
In-Vivo Estimates of Axonal Characteristics Using Optimized Diffusion MRI Protocols for Single Fibre Orientation (2010) — contributed 3 concept elements to this direction.