Abstract
ObjectiveThe purpose of this study was to utilize a convolutional neural network (CNN) to make preoperative differential diagnoses between ameloblastoma (AME) and odontogenic keratocyst (OKC) on cone-beam CT (CBCT).MethodsThe CBCT images of 178 AMEs and 172 OKCs were retrospectively retrieved from the Hospital of Stomatology, Wuhan University. The datasets were randomly split into a training dataset of 272 cases and a testing dataset of 78 cases. Slices comprising lesions were retained and then cropped to suitable patches for training. The Inception v3 deep learning algorithm was utilized, and its diagnostic performance was compared with that of oral and maxillofacial surgeons.ResultsThe sensitivity, specificity, accuracy, and F1 score were 87.2%, 82.1%, 84.6%, and 85.0%, respectively. Furthermore, the average scores of the same indexes for 7 senior oral and maxillofacial surgeons were 60.0%, 71.4%, 65.7%, and 63.6%, respectively, and those of 30 junior oral and maxillofacial surgeons were 63.9%, 53.2%, 58.5%, and 60.7%, respectively.ConclusionThe deep learning model was able to differentiate these two lesions with better diagnostic accuracy than clinical surgeons. The results indicate that the CNN may provide assistance for clinical diagnosis, especially for inexperienced surgeons.
Highlights
Ameloblastoma (AME) and odontogenic keratocyst (OKC) are common radiolucent lesions of the jaws in oral and maxillofacial surgery [1, 2]
The differentiation between AME and OKC in radiography is mainly based on some features, such as buccolingual expansion, the number of locules, internal density, and the root resorption of the adjacent teeth (Figure 1)
Inception v3 obtained the highest scores among the participants, with a sensitivity of 87.4%, a specificity of 82.1%, an accuracy of 84.6%, and an F1 score of 85.0% (Table 2)
Summary
Ameloblastoma (AME) and odontogenic keratocyst (OKC) are common radiolucent lesions of the jaws in oral and maxillofacial surgery [1, 2]. Current treatment modalities for AME are wide local excision and immediate reconstruction, but OKC is generally treated with more conservative surgical methods, such as marsupialization and/or enucleation Given that they have different treatment strategies, it is imperative to differentiate these conditions before surgery [5,6,7,8]. Previous studies have sought more instrumental radiographic findings, such as the width-to-length ratio, volumetric measurement, and assessment of the Hounsfield unit, to distinguish these two lesions [9,10,11]. These studies have the same limitation in that they only focused on low-level and limited features. It can be contended that the current knowledge of radiography is still at tip of the iceberg, and more undetected information waits to be mined
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
