A Combined Radiomics and Anatomical Features Model Enhances MRI-Based Recognition of Symptomatic Nerves in Primary Trigeminal Neuralgia

Motivation: This study addresses radiologists' limited accuracy in diagnosing primary trigeminal neuralgia (PTN) and explores its pathology through radiomics. Goal(s): To create a PTN diagnostic model combining radiomic and clinical features, demonstrating that neurovascular compression (NVC) isn't the primary cause of PTN. Approach: Data from 279 trigeminal nerves in 140 PTN patients were retrospectively analyzed. Regions of interest (ROI) in the trigeminal cisternal segment and Meckel cavity (MC) provided 2632 radiomics features; 13 clinical features were included. Lasso regression and other methods were used. Results: The model with two ROIs and clinical features achieved an AUC of 0.91/0.90, surpassing radiologists in identifying symptomatic nerves. Impact: This study has greatly improved the accuracy of radiologists in diagnosing the disease, which is conducive to the early detection and treatment of the disease; It also provides the basis for the study of the pathophysiological mechanism of PTN.