Abstract
OBJECTIVES: This study aims to characterise the rare 'fainting wisdom tooth' phenomenon, defined as the angular transition of a mandibular wisdom tooth from an initially upright position to mesial or horizontal inclination during adolescence in fully dentate mandibles. MATERIALS AND METHODS: A retrospective longitudinal analysis was conducted using panoramic radiographs from a pooled database of 11,932 patients. Fifty cases of fainting wisdom teeth (0.4% prevalence) were identified using a validated deep learning (DL) convolutional neural network (CNN)-based artificial intelligence (AI) tool and compared to 50 matched controls with vertically erupted wisdom teeth. Radiographic variables were measured at two time points: T1 (ages 8-15) and T2 (ages 16-23). Angular changes and anatomical features, including space/crown width ratio, tooth depth, proximity to the second molar and mandibular canal, cemento-enamel junction (CEJ)/apical width ratio and gonial angle, were assessed. Logistic regression was used to determine predictive factors associated with the fainting phenomenon, with results reported as odds ratios (ORs). RESULTS: Fainting wisdom teeth showed a significant angular shift from a median of 29° at T1 to 83° at T2, while controls showed minimal change. Logistic regression identified tooth depth as the strongest predictor (OR = 0.1, p < 0.001), with additional risk factors including reduced space/crown width ratio, increased proximity to the second molar, higher CEJ/apical width ratio and increasing gonial angle. CONCLUSION: This is the first study to describe and analyse the rare fainting wisdom tooth phenomenon using AI-assisted longitudinal analysis of panoramic radiographs. The findings suggest that early tooth positioning and surrounding anatomical constraints may interfere with vertical eruption and increase the risk of fainting.