MRI Reveals Hidden Architecture Inside Plant-Based Meat

The fibrous, meat-like bite of a plant-based burger is not an accident of chemistry but the product of an elaborate structural hierarchy that has, until now, been largely invisible to food scientists. In a study published in Current Research in Food Science, a team of Dutch researchers led by Sam Kuijpers and Camilla Terenzi has turned to magnetic resonance imaging, the same family of techniques used to map the human brain, to peer inside high-moisture extrudates made from soy protein concentrate. Their work reveals, for the first time in a single coherent framework, how structures at scales spanning from a tenth of a micrometre to hundreds of micrometres conspire to produce the anisotropic, layered texture that makes extruded plant protein feel like meat on the palate.

High-moisture extrusion is the industrial workhorse behind most plant-based meat alternatives. A protein dough, typically derived from soy or pea isolates, is hydrated, mixed and then pushed through a heated barrel where temperatures reach 135 degrees Celsius and intense shear forces from co-rotating screws knead the material. The molten protein then passes through a cooling die, where it solidifies into a ribbon with a lamellar, fibrous structure reminiscent of muscle tissue. The critical question has always been how this hierarchical anisotropy forms: which processes happen in the barrel, which in the cooling die, and how events at the nanometre scale cascade upward to determine the texture a consumer actually perceives when chewing.

To answer this, the researchers deployed an unusually broad arsenal of magnetic resonance methods. T2-weighted MRI at 14 tesla visualised the phase-separated architecture, distinguishing protein-rich domains from water-rich lamellae at the hundred-micrometre scale. Diffusion tensor MRI, a technique borrowed from clinical neuroimaging where it maps white matter tracts, was used to measure the directionality of water self-diffusion within each voxel, revealing the orientation of structures far below the imaging resolution. Chemical exchange saturation transfer MRI, performed at the extraordinary field strength of 28.2 tesla, probed the chemical composition of the two phases, detecting mobile proteins dissolved in the water-rich regions through their amide and relayed nuclear Overhauser enhancement signatures.

The team extruded commercial soy protein concentrate at 60 percent moisture and shifted the pH of the feed liquid to three values: acidic pH 5.0, near the isoelectric point of the protein; neutral pH 7.0; and alkaline pH 7.8. The results were striking. At neutral and alkaline pH, the extrudates displayed beautifully aligned lamellar structures running parallel to the extrusion direction, with weighted order parameters reaching 0.9 in sagittal views. At pH 5.0, by contrast, the macroscopic lamellar order collapsed almost entirely, with order parameters dropping to around 0.4 and no uniform aligned phase separation visible along the flow direction.

Yet the story at smaller length scales was inverted. Confocal laser scanning microscopy, capable of resolving features down to roughly 100 nanometres, showed that the acidic sample actually possessed a well-aligned fibrillar protein network at the sub-micrometre scale, with order parameters around 0.6, comparable to the neutral sample. The alkaline sample, meanwhile, lacked this fine-scale fibrillar alignment altogether, a consequence of strong electrostatic repulsion between highly charged protein fibrils far from their isoelectric point. This decoupling between order at different scales provided the central clue to the mechanism of structure formation.

The diffusion tensor measurements added quantitative weight to this picture. In the protein-rich lamellae of the pH 7.0 sample, water diffused preferentially along the extrusion direction, the fastest diffusion direction coinciding with the shear imposed by the extruder screws, while the slowest diffusion ran perpendicular to the lamellae. Fractional anisotropy values of about 0.3 in the protein-rich phase, against 0.2 in the water-rich phase, confirmed that the aligned protein fibrils observed microscopically were genuinely hindering water motion in a directional manner. Because the diffusion encoding time of 50 milliseconds probes water displacement over roughly 9 to 33 micrometres, the technique effectively senses the cumulative effect of countless sub-micrometre fibrils, bridging the gap between what microscopy and conventional MRI can each see.

To understand why the cooling die geometry matters, the researchers generated synthetic axial MRI images mimicking lamellar patterns for four die cross-sections: elliptical, bullnose, rounded-rectangle and sharp-cornered rectangular. The rotated Fourier transform analysis, a recently developed image-processing method that extracts orientational order from two-dimensional images, showed that elliptical dies produce highly homogeneous order parameters near 0.93, while progressively sharper corners carve out increasingly wide zones of disorder, with order parameters falling to 0.3 in the corners of the sharpest geometry. The experimental maps of real extrudates matched the sharp-cornered simulation almost exactly, with a disordered corner region roughly 1.5 millimetres wide, demonstrating that the die’s shape directly imprints its geometry onto the internal texture of the product.

Synthesising these observations, the authors propose a two-stage mechanism. Fibril formation and alignment at the sub-micrometre scale begin already in the extruder barrel and transition zone. The degree of this fine-scale alignment, governed by pH through electrostatic screening, then controls the rheology of the protein melt entering the cooling die. At pH 5.0, the strongly aligned fibrillar network raises the yield stress so much that the melt can no longer flow in laminar sheets through the die, and the macroscopic lamellar structure fails to form. At pH 7.8, weak fibril alignment keeps the yield stress low, allowing coarse lamellae to form but without the underlying fibrillar backbone. Only at an intermediate pH, where fibrillar alignment and electrostatic repulsion strike a balance, do both levels of hierarchy emerge in concert. As a compelling validation of the method’s specificity, the team showed that commercial high-moisture mozzarella, whose casein structure is aligned only by stretching at much lower temperatures, exhibits order at the hundred-micrometre scale but no diffusional anisotropy at smaller scales, consistent with small-angle neutron scattering showing isotropy below 90 nanometres.

The implications reach well beyond academic curiosity. Plant-based meat alternatives are a cornerstone of the protein transition, yet their texture remains the chief barrier to consumer acceptance. By establishing high-resolution MRI, diffusion tensor imaging and CEST as a versatile, non-destructive toolkit for quantifying hierarchical anisotropy in intact extrudates, the study gives food engineers, for the first time, a way to watch structure form across six orders of magnitude in length scale and to rationally tune pH, die geometry and processing conditions. The work was funded by the Dutch Research Council together with industrial partners including Unilever, Cargill, dsm-firmenich and FrieslandCampina, underscoring the commercial stakes. If the meat-like fibrousness of the next generation of plant-based products improves noticeably, it may well be because a scanner originally built for brains taught manufacturers how to build better fibres.

Subject of Research: Magnetic resonance imaging of hierarchical anisotropic structure formation in high-moisture soy protein extrudates

Article Title: Underpinning hierarchical anisotropic structure formation in soy protein high-moisture extrudates by Magnetic Resonance Imaging

Article References: Kuijpers, S. A., Gobes, M. I., Mayar, M., Vergeldt, F. J., Hohlbein, J., Huppertz, T., van Duynhoven, J. P., & Terenzi, C. (2026). Underpinning hierarchical anisotropic structure formation in soy protein high-moisture extrudates by Magnetic Resonance Imaging. Current Research in Food Science, 13, Article 101575. https://doi.org/10.1016/j.crfs.2026.101575

Image Credits: AI Generated

DOI: Not provided

Keywords: plant-based meat, high-moisture extrusion, soy protein, MRI, diffusion tensor imaging, CEST, food structure, anisotropy, pH-shifting, cooling die geometry, confocal microscopy, food science