Use of an Artificial Neural Network for Diagnosis of Facial Pain Syndromes: An Update

Background: Based on a classification scheme for facial pain syndromes and a binomial (yes/no) facial pain questionnaire, we previously reported on the ability of an artificial neural network (ANN) to recognize and correctly diagnose patients with different facial pain syndromes. Objectives: We now report on an updated questionnaire, the development of a secure web-based neural network application and details of ANNs trained to diagnose patients with different facial pain syndromes. Methods: Online facial pain questionnaire responses collected from 607 facial pain patients (395 female, 65%, ratio F/M 1.86/1) over 5 years and 7 months were used for ANN training. Results: Sensitivity and specificity of the currently running ANN for trigeminal neuralgia type 1 and trigeminal neuralgia type 2 are 92.4 and 62.5% and 87.8 and 96.4%, respectively. Sensitivity and specificity are 86.7 and 95.2% for trigeminal neuropathic pain, 0 and 100% for trigeminal deafferentation pain and 100% for symptomatic trigeminal neuralgia and postherpetic neuralgia. Sensitivity is 50% for nervus intermedius neuralgia (NIN) and 0% for atypical facial pain (AFP), glossopharyngeal neuralgia (GPN) and temporomandibular joint disorder (TMJ). Specificity for AFP, NIN and TMJ is 99% and for GPN, 100%. Conclusions: We demonstrate the utilization of question-based historical self-assessment responses used as inputs to design an ANN for the purpose of diagnosing facial pain syndromes (outputs) with high accuracy.