3D nerve shape outperformed conventional MRI metrics in trigeminal neuralgia

In trigeminal neuralgia, the anatomy of the nerve has always mattered, but not every compressed nerve hurts in the same way. This preprint argues that the gap between compression and pain may be visible in the nerve’s shape itself, if the anatomy is measured with enough resolution and the right mathematical tools.

Investigators analyzed a cohort of 479 patients with trigeminal neuralgia who underwent microvascular decompression and had high-resolution pre-operative MRI scans. The central question was straightforward: can data-driven 3D shape features capture pain-associated morphology better than the hand-crafted geometric measurements that have been used for years? In a held-out test cohort, the answer was yes. A predictive model built from the data-driven shape features achieved an AUROC of 0.719, compared with 0.626 for a model using existing predefined measurements, with p=0.005.

The difference is not cosmetic. It suggests that the biologic signal linked to pain may live in higher-order anatomy that conventional metrics flatten away.

Why compression alone has never fully explained pain laterality

The leading model for trigeminal neuralgia remains neurovascular compression, yet compression by itself does not determine which nerves are painful. That clinical mismatch has long hinted that the relevant lesion may not be simple contact alone, but a broader morphologic response: distortion, remodeling, and subtle deformation of the nerve as it travels through the cerebellopontine angle. A nerve that is squeezed, stretched, or repeatedly pulsated against a vessel may change shape in ways that are difficult to appreciate with standard measurements.

This study treats those changes as a data problem rather than a purely visual one. Instead of depending on predefined distances, angles, or thickness measures, the team extracted 3D shape features designed to preserve the geometry of the trigeminal nerve more completely. That approach matters because pain may emerge from combinations of deformation patterns rather than from one isolated dimension. A small change in curvature, tapering, or regional asymmetry can be clinically invisible to the eye and yet still reflect chronic mechanical stress on the nerve.

Higher-order morphology may be the biologic footprint of painful nerves

The study’s main contribution is not only a statistical improvement, but a conceptual shift in how nerve injury is represented. Conventional hand-crafted measurements are useful when anatomy changes in simple, linear ways. They become less informative when disease produces irregular, nonuniform patterns across the full nerve surface or along multiple segments at once.

Data-driven shape modeling is better suited to that problem. By learning from the geometry itself, the model can detect combinations of features that act together as a biomarker of pain-associated morphology. In practical terms, that could mean the painful nerve is not merely narrower, larger, or more displaced, but structurally different in a way that only emerges when the entire 3D form is analyzed as a single object. The reported AUROC gain from 0.626 to 0.719 implies that the higher-order descriptors captured clinically relevant information that older measurements missed.

For surgeons and imaging specialists, that distinction points toward a more refined preoperative phenotype. The painful nerve may carry a recognizable structural signature before it is ever exposed in the operating room.

What the cohort and model can, and cannot, establish

The dataset was substantial for a morphologic imaging study, with 479 patients and postoperative confirmation in the setting of microvascular decompression. Even so, the work remains a preprint and has not undergone peer review. That matters because model performance can shift when methods are validated across institutions, scanners, and segmentation pipelines.

The study also does not claim that shape alone explains pain. It shows association, not mechanism. Neurovascular compression likely remains part of the causal chain, but the data suggest that the nerve’s structural response to that compression may be the more proximate marker of symptoms. A biomarker built from MRI shape features may therefore function as a surrogate for chronic mechanical injury, local remodeling, or both. None of those possibilities can be separated from this dataset alone.

Still, the result is clinically provocative. A model that better distinguishes painful from nonpainful nerves could eventually support preoperative counseling, operative planning, and hypothesis generation about which anatomical patterns are most likely to matter in symptom production.

Imaging biomarkers may eventually refine patient selection for surgery

Microvascular decompression is a procedure built around anatomy, and anatomy is often the part of the case that appears deceptively simple. Patients with trigeminal neuralgia may have visible vessel-nerve contact, yet not every contact explains their symptoms with equal force. If shape-based biomarkers continue to outperform traditional metrics, high-resolution MRI could move beyond anatomic confirmation and toward a more discriminating read on disease biology.

That would be especially valuable in a disorder where the clinical stakes are high and the therapeutic threshold is narrow. Trigeminal neuralgia can be profoundly disabling, and when surgery is considered, clinicians need more than a binary yes-or-no answer about compression. They need a way to identify the nerve most likely to be driving the pain. A morphology-based model could help separate incidental contact from contact that has already left a measurable imprint on the nerve.

The investigators’ approach also fits a broader trend in neuroimaging: instead of relying solely on human inspection, use machine learning and shape analysis to quantify subtle structural disease signatures. In this case, the target was a cranial nerve, but the logic extends to other conditions where anatomy carries disease information that is hard to express with standard measurements.

The preprint leaves several questions open, including how well the model performs in external cohorts, whether the same shape features generalize across MRI protocols, and whether the biomarkers correlate with postoperative outcome or pain severity. Those next steps will determine whether trigeminal nerve shape becomes a research signal only, or a clinically useful adjunct to the imaging workup of trigeminal neuralgia.


Reporting by Dr. A. Manasa, MBBS for Medicine World Foundation, with additional data provided by medRxiv (Clinical Preprints).
PRIMARY CLINICAL EVIDENCE & REGISTRY DATA