IADR Abstract Archives
An artificial intelligence model for the radiographic diagnosis of osteoarthritis of the temporomandibular joint
Objectives: Osteoarthritis is a chronic inflammatory disease that has serious consequences affecting the quality of life. The diagnosis of temporomandibular disorders (TMD) often suffers from the inconsistency and controversies that surrounds the diagnostic protocols. This study aimed to develop and test the performance of an AI model based on neural networks for the diagnosis of TMJ osteoarthritis from CBCT.
Methods: Patients were assessed using the DC/TMD Axis-I and Axis-II assessment instruments. The radiographic criteria that were used to confirm the diagnosis of osteoarthritis were flattening of the articular surfaces, osteophyte, subcortical cyst and surface erosion. A total of 2737 images were used for the training and validation of the AI model. The odds of the correct CBCT diagnosis were then compared between the AI model and one oral radiologist against a golden reference. The model used a single convolutional network to concurrently predict several bounding boxes and generate the class probabilities. A separate testing set composed of 321 images was used for testing. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), test accuracy and the Kappa coefficient of agreement were used to evaluate the performance of the AI model.
Results: Both diagnostic methods showed near perfect agreement with the golden reference. AI Diagnosis showed better agreement with golden reference compared to the examiner diagnosis. Osteophyte showed the lowest agreement whereas erosion showed the largest agreement with the golden reference. AI showed statistically highly significant agreement with golden reference compared to the examiner (P=0.0000). The Kappa coefficient of agreement showed near perfect agreement between the AI model and the golden reference
Conclusions: The diagnosis of TMJ osteoarthritis from CBCT was possible using neural network. This may eliminate the subjectivity and lead to early diagnosis of this progressive disorder.
Methods: Patients were assessed using the DC/TMD Axis-I and Axis-II assessment instruments. The radiographic criteria that were used to confirm the diagnosis of osteoarthritis were flattening of the articular surfaces, osteophyte, subcortical cyst and surface erosion. A total of 2737 images were used for the training and validation of the AI model. The odds of the correct CBCT diagnosis were then compared between the AI model and one oral radiologist against a golden reference. The model used a single convolutional network to concurrently predict several bounding boxes and generate the class probabilities. A separate testing set composed of 321 images was used for testing. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), test accuracy and the Kappa coefficient of agreement were used to evaluate the performance of the AI model.
Results: Both diagnostic methods showed near perfect agreement with the golden reference. AI Diagnosis showed better agreement with golden reference compared to the examiner diagnosis. Osteophyte showed the lowest agreement whereas erosion showed the largest agreement with the golden reference. AI showed statistically highly significant agreement with golden reference compared to the examiner (P=0.0000). The Kappa coefficient of agreement showed near perfect agreement between the AI model and the golden reference
Conclusions: The diagnosis of TMJ osteoarthritis from CBCT was possible using neural network. This may eliminate the subjectivity and lead to early diagnosis of this progressive disorder.