Artificial intelligence in trigeminal neuralgia: trigeminal nerve segmentation and neurovascular conflict detection: a systematic review and meta-analysis

The study is the first systematic review and meta-analysis to pool the diagnostic performance of AI models for trigeminal nerve segmentation and neurovascular conflict detection in trigeminal neuralgia diagnosis. Whereas prior work was mostly individual studies, this review systematically searched PubMed, Scopus, Embase, Web of Science, and the Cochrane Library and pooled data from 5 studies and 577 confirmed patients to produce combined estimates of sensitivity, specificity, likelihood ratios, diagnostic odds ratio, and AUC. Based on 5 studies meeting inclusion criteria and 577 patients with confirmed trigeminal neuralgia, with meta-analysis of pooled sensitivity, specificity, positive/negative diagnostic likelihood ratios, diagnostic odds ratio, and AUC, and extraction of reference standards including expert imaging-based annotation, clinical-radiological diagnosis, and intraoperative confirmation.

AI models showed high specificity and a high diagnostic odds ratio for detecting trigeminal neuralgia, with pooled specificity of 92% and diagnostic odds ratio of 26.71. This quantifies the performance of AI models in ruling out non-trigeminal-neuralgia cases, with a positive diagnostic likelihood ratio of 8.5 and a negative diagnostic likelihood ratio of 0.32, providing comparable statistical grounds for clinical decision-making. Pooled specificity 92% (95% CI: 88-94%), diagnostic odds ratio 26.71 (95% CI: 16.38-43.56), positive likelihood ratio 8.5 (95% CI: 5.75-12.56), and negative likelihood ratio 0.32 (95% CI: 0.26-0.38), all from the meta-analysis of 5 studies.

Deep learning and machine learning approaches, particularly deep learning, show promising diagnostic performance in trigeminal nerve segmentation and neurovascular conflict detection and can aid preoperative planning. The study links the ability of AI to automatically segment the trigeminal nerve and identify neurovascular conflict on MRI to the clinical diagnostic workflow, noting it can help improve trigeminal neuralgia diagnostic accuracy and preoperative planning for microvascular decompression. Pooled AUC reached 0.91 (95% CI: 0.88-0.93) and sensitivity was 71% (95% CI: 65-76%), with the studies focused on automated segmentation and neurovascular conflict detection in MRI scans.

The results apply to settings where AI models are used for trigeminal nerve segmentation and neurovascular conflict detection on MRI scans, in patients with confirmed trigeminal neuralgia being considered for treatments such as microvascular decompression; the evidence comes from 5 studies and 577 patients, with reference standards including expert imaging-based annotation, clinical-radiological diagnosis, and intraoperative confirmation.

Only 5 studies and 577 patients were included, so the confidence intervals of the pooled estimates and sources of heterogeneity warrant attention; reference standards across studies include expert imaging-based annotation, clinical-radiological diagnosis, and intraoperative confirmation, which may affect interpretation; whether AI performance in trigeminal nerve segmentation and neurovascular conflict detection generalizes to broader populations and different imaging workflows still requires further study.