AI has advanced significantly across various medical disciplines, particularly in dentistry, where it enhances the diagnosis, localization, classification, estimation, and assessment of dental diseases. Recent advancements in AI technologies tailored for dental practitioners enable precise diagnoses and accurate treatment recommendations. The following figure illustrates the specific areas within dentistry where AI applications are deployed, accompanied by detailed descriptions of their practical implementations. In this newsletter we want to focus on the application of AI in Periodontology.
Periodontology is concerned with the prevention, diagnosis, and treatment of diseases that affect the supporting structures of teeth. These structures include the gums (gingiva), alveolar bone, periodontal ligament, and cementum, collectively known as the periodontium. Periodontal diseases, such as gingivitis and periodontitis, involve inflammation and infections that can lead to progressive damage if untreated. Effective periodontal care is essential not only for maintaining oral health but also for preventing systemic health issues associated with untreated periodontal disease.
AI has profoundly impacted periodontology through advancements in various applications, including the detection of periodontal bone loss, diagnosis of gingivitis inflammation, and assessment of connective tissues and other periodontal conditions. To augment diagnostic accuracy, researchers have employed machine learning techniques. Li et al. pioneered a plaque segmentation method using convolutional neural networks (CNN) on oral endoscopic images, achieving an impressive accuracy of 86.42%. Another study by Li et al. utilized an extreme machine learning approach with digital photographs, achieving 74% accuracy in identifying gingivitis. Lin et al. explored a level segmentation method for localizing alveolar bone loss, demonstrating robust efficacy through support vector machines (SVM), k-nearest neighbor (KNN), and Bayesian classifiers.
Deep learning (DL) methods have garnered attention, particularly in detecting bone loss from intraoral radiographs. Lee et al. employed a VGG-based neural network to diagnose periodontal bone loss with 99% accuracy and 98% area under the curve (AUC), surpassing the performance of three dentists. Krois et al. utilized a deep-feed forward CNN on panoramic radiographs, achieving comparable discrimination ability to three examiners. Transfer learning techniques, as suggested by Kim et al. and Lee et al., enhanced bone loss and odontogenic cyst lesion detection using panoramic radiographs, yielding superior results in tooth numbering compared to dental clinicians. Moran et al. demonstrated the effectiveness of a ResNet model in classifying regions based on periodontal bone destruction with 82% accuracy from periapical radiographs.
Automation in identifying bone lesions and shapes has been advanced by Khan et al. through a disease segmentation method based on U-Net architecture, surpassing expert dentists in caries detection. Zheng et al. proposed an anatomically constrained dense U-Net for identifying bone lesions using cone beam computed tomography (CBCT), accurately delineating bone shapes and lesions. Duong et al. developed a U-Net-based network for alveolar bone delineation from high-frequency ultrasound images, achieving performance exceeding that of expert evaluations. Nguyen et al. utilized ResNet with U-Net for alveolar bone segmentation from intraoral images, achieving a dice coefficient of 85.3%. For assessing periodontitis severity, Li et al. employed Mask R-CNN on panoramic radiographs, achieving 82% accuracy, surpassing junior dentists in diagnostic precision.
In dentistry, accurate data interpretation and diagnosis are pivotal. AI systems, categorized as Clinical Decision Support Systems (CDSS), play a crucial role in facilitating precise medical decisions in time-sensitive environments. Artificial Neural Networks (ANNs) have been pivotal in this regard. Geetha et al. proposed a backpropagation neural network for detecting tooth decay from intraoral radiographs, achieving 97.1% accuracy and a low false positive rate of 2.8%. Papantonopoulos et al. evaluated a multilayer perceptron ANN for assessing bone loss from medical health records, achieving effective periodontitis classification with 98.1% accuracy. Shankarapillai et al. developed a multilayer feed-forward propagation network for predicting periodontitis risk, demonstrating its efficacy with textual data from 230 subjects.
In conclusion, Artificial Intelligence (AI) has ushered in a transformative era in periodontology, significantly enhancing diagnostic capabilities and treatment outcomes across various domains of dental care. The application of AI technologies, including machine learning and deep learning models like convolutional neural networks (CNNs) and U-Net architectures, has revolutionized tasks such as periodontal bone loss detection, gingivitis diagnosis, and disease assessment with unprecedented accuracy and efficiency. These advancements not only surpass traditional methods but also empower dental professionals with precise tools for personalized treatment planning and patient care. AI-driven Clinical Decision Support Systems (CDSS) have further bolstered clinical decision-making by providing timely, data-driven insights that improve diagnostic accuracy and treatment efficacy. Moving forward, the ongoing evolution and integration of AI in periodontology promise to continue driving innovation, optimizing workflows, and elevating standards in oral healthcare delivery worldwide.
