Classifier Using Pontine Radial Diffusivity and Symptom Duration Accurately Predicts Recurrence of Trigeminal Neuralgia After Microvascular Decompression: A Pilot Study and Algorithm Description.

Abstract

Preprocedure diffusion tensor magnetic resonance imaging (MRI) may predict the response of trigeminal neuralgia (TN) patients to Gamma Knife (Elekta AB) and microvascular decompression (MVD). To test this hypothesis using pontine-segment diffusion tensor MRI radial diffusivity (RD), a known biomarker for demyelination, to predict TN recurrence following MVD. RD from the pontine segment of the trigeminal tract was extracted in a semiautomated and blinded fashion and normalized to background pontine RD. Following validation against published results, the relationship of normalized RD to symptom duration (DS) was measured. Both parameters were then introduced into machine-learning classifiers to group patient outcomes as TN remission or recurrence. Performance was evaluated in an observational study with leave-one-out cross-validation to calculate accuracy, sensitivity, specificity, and receiver operating characteristic curves. The study population included 22 patients with TN type 1 (TN1). There was a negative correlation of normalized RD and preoperative symptom duration (P=.035, R2=.20). When pontine-segment RD and DS were included as input variables, 2 classifiers predicted pain-free remission versus eventual recurrence with 85% accuracy, 83% sensitivity, and 86% specificity (leave-one-out cross-validation; P=.029) in a cohort of 13 patients undergoing MVD. Pontine-segment RD and DS accurately predict MVD outcomes in TN1 and provide further evidence that diffusion tensor MRI contains prognostic information. Use of a classifier may allow more accurate risk stratification for neurosurgeons and patients considering MVD as a treatment for TN1. These findings provide further insight into the relationship of pontine microstructure, represented by RD, and the pathophysiology of TN.