Unanticipated difficult laryngoscopy is associated with serious airway-related complications. We aimed to develop and test a convolutional neural network-based deep-learning model that uses lateral cervical spine radiographs to predict Cormack_ehane grade 3 or 4 direct laryngoscopy views of the glottis. We analysed the radiographs of 5939 thyroid surgery patients at our hospital, 253 (4%) of whom had grade 3 or 4 glottic views. We used 10 randomly sampled datasets to train a model. We compared the new model with six similar models (VGG, ResNet, Xception, ResNext, DenseNet and SENet). The Brier score (95%CI) of the new model, 0.023 (0.0210.025), was lower (_etter) than the other models- VGG, 0.034 (0.0340.035); ResNet, 0.033 (0.0330.035); Xception, 0.032 (0.0310.033); ResNext, 0.033 (0.0320.033); DenseNet, 0.030 (0.0290.032); SENet, 0.031 (0.0290.032), all p<0.001. We calculated mean (95%CI) of the new model for- R2, 0.428 (0.3880.468); mean squared error, 0.023 (0.0210.025); mean absolute error, 0.048 (0.0460.049); balanced accuracy, 0.713 (0.6840.742); and area under the receiver operating characteristic curve, 0.965 (0.9620.969). Radiographic features around the hyoid bone, pharynx and cervical spine were associated with grade 3 and 4 glottic views.