Artificial Intelligence in Trigeminal Neuralgia: Trigeminal Nerve Segmentation and Neurovascular Conflict Detection: A Systematic Review and Meta-Analysis Publisher Pubmed



Alvani M S ; Mohammadzadeh I ; Hajikarimloo B ; Javani M ; Mohammadzadeh S ; Mortezaei A ; Zare A ; Fayedeh F ; Sanikommu S ; Khalaji A ; Habibi M A ; Dmytriw A A ; Fahim F ; Mousavinejad S A Show All Authors
Source: Neurosurgical Review Published:2026

Abstract

Trigeminal neuralgia (TN) is a debilitating condition characterized by severe, episodic facial pain, often caused by neurovascular conflict (NVC) at the trigeminal nerve root. Early and accurate diagnosis of TN is crucial for effective management, with microvascular decompression (MVD) being a common surgical treatment. While MRI plays a key role in detecting NVC, its subjective nature limits diagnostic precision. Recent advances in artificial intelligence (AI) offer promising tools for enhancing diagnostic accuracy. This study evaluates the potential of deep learning (DL) and machine learning (ML) models in improving TN diagnosis through automated segmentation and NVC detection in MRI scans. A comprehensive search was conducted across PubMed, Scopus, Embase, Web of Science, and the Cochrane Library to identify studies utilizing AI models in TN diagnosis. Studies assessing sensitivity, specificity, accuracy, and AUC were included in the meta-analysis. A total of 5 studies met the inclusion criteria, encompassing 577 patients with confirmed TN. The primary focus was on evaluating the performance of AI models in segmenting the trigeminal nerve and identifying neurovascular conflicts. The reference standard or target criterion used to define correct model performance was also extracted and considered, including expert imaging-based annotation, clinical-radiological diagnosis, and intraoperative confirmation where available. The pooled sensitivity of AI models in detecting TN was 71% (95% CI: 65–76%), with a specificity of 92% (95% CI: 88–94%). The positive diagnostic likelihood ratio (DLR) was 8.5 (95% CI: 5.75–12.56), and the negative DLR was 0.32 (95% CI: 0.26–0.38). The diagnostic odds ratio (DOR) was 26.71 (95% CI: 16.38–43.56), and the area under the curve (AUC) reached 0.91 (95% CI: 0.88–0.93). These findings demonstrate the high diagnostic accuracy of AI models, particularly in identifying NVC and improving preoperative planning for TN patients. AI models, particularly DL-based approaches, show promising diagnostic performance in the segmentation of the trigeminal nerve and detection of NVC. The high specificity and diagnostic odds ratio suggest that AI can play a critical role in enhancing clinical decision-making and improving the accuracy of TN diagnosis, ultimately aiding in better treatment planning for patients. © The Author(s) 2026.