A novel artificial neural network for the diagnosis of orofacial pain and temporomandibular disorders

Background Temporomandibular disorders (TMD) and orofacial pain are highly prevalent. This prevalence can be compared to that of leading non‐communicable diseases (NCDs). However, it is surprising to still find a high degree of controversy regarding its diagnosis and management. Patients usually experience treatment delays, missed diagnoses, and receive unnecessary therapies. New artificial intelligence algorithms have helped diagnose numerous diseases. Nevertheless, no studies have focused on the use of artificial intelligence to diagnose these conditions. Objectives This study aimed to develop and test the performance of a novel neural network (multilayer perceptron) with diagnostic capabilities in orofacial pain and TMD, including some types of referred pain. Methods A multilayer perceptron (MLP) was developed with one input layer, five hidden layers, and one output layer. It was trained using backpropagation algorithms. Several categories of orofacial pain and TMD clinical cases were presented to 12 general dental clinicians, and their diagnoses were contrasted to those provided by the artificial intelligence neural network. Results Overall, the diagnostic accuracy of the artificial intelligence was superior to that of the general dental clinicians (p = .0072). This was more evident in the clinical cases involving non‐dental and referred orofacial pains (e.g. neuropathic pain, referred cardiac pain, neurovascular pain). Conclusions This study showed, for the first time, that an artificial neural network can help medical and general dental clinicians diagnose several types of orofacial pain and dysfunction, including TMD, neuropathic, neurovascular, and referred cardiac pain. In some cases, the MLP appears to have a life‐saving role. A novel artificial intelligence system (multilayer perceptron) with diagnostic capabilities in orofacial pain and TMD was developed and tested. The multilayer perceptron was constructed with 1 input layer of 18 artificial neurons, 5 hidden layers, and 1 output layer. Its diagnostic capability was contrasted to those provided by general dental clinicians. Overall, the diagnostic accuracy of the artificial intelligence was superior to that of the general dental clinicians (p = .0072).