AIM: Atherosclerotic carotid plaques (ACPs) constitute the main etiological factor in about 15% of strokes. ACPs can be detected on routine dental panoramic radiographs. As these are one of the most commonly performed dental images, they can be used as a source of available data for computerized methods of automatic detection of ACPs in order to significantly increase their timely diagnosis. The aim of this study was to present the potential of applying deep learning methodology to detect ACPs on routine panoramic radiographs with the ultimate goal of preventing strokes. METHODS: The Faster Region-based Convolutional Neural Network (Faster R-CNN) for deep learning was used. The operation of the algorithm was assessed on a small dataset of 65 panoramic images. As the available training data was limited, data augmentation was performed by changing the brightness and randomly flipping and rotating cropped regions of interest in multiple angles. Receiver operating characteristic (ROC) analysis was performed to calculate the accuracy of detection. RESULTS: ACPs were detected with a sensitivity of 75%, a specificity of 80%, and an accuracy of 83%. The ROC analysis showed a significant area under curve (AUC), different from 0.5. CONCLUSIONS: The novelty of the study lies in showing the efficiency of a deep learning method for the detection of ACPs on routine panoramic images based on a small dataset. Further improvement is needed as regards the application of the algorithm to the level of introducing this methodology in routine dental practice for stroke prevention.
AIM: Atherosclerotic carotid plaques (ACPs) constitute the main etiological factor in about 15% of strokes. ACPs can be detected on routine dental panoramic radiographs. As these are one of the most commonly performed dental images, they can be used as a source of available data for computerized methods of automatic detection of ACPs in order to significantly increase their timely diagnosis. The aim of this study was to present the potential of applying deep learning methodology to detect ACPs on routine panoramic radiographs with the ultimate goal of preventing strokes. METHODS: The Faster Region-based Convolutional Neural Network (Faster R-CNN) for deep learning was used. The operation of the algorithm was assessed on a small dataset of 65 panoramic images. As the available training data was limited, data augmentation was performed by changing the brightness and randomly flipping and rotating cropped regions of interest in multiple angles. Receiver operating characteristic (ROC) analysis was performed to calculate the accuracy of detection. RESULTS: ACPs were detected with a sensitivity of 75%, a specificity of 80%, and an accuracy of 83%. The ROC analysis showed a significant area under curve (AUC), different from 0.5. CONCLUSIONS: The novelty of the study lies in showing the efficiency of a deep learning method for the detection of ACPs on routine panoramic images based on a small dataset. Further improvement is needed as regards the application of the algorithm to the level of introducing this methodology in routine dental practice for stroke prevention.