Deep Learning Applications in Dental Image-Based Diagnostics: A Systematic Review

Background: AI has been adopted in dentistry for diagnosis, decision making, and therapy prognosis prediction. This systematic review aimed to identify AI models in dentistry, assess their performance, identify their shortcomings, and discuss their potential for adoption and integration in dental practice in the future. Methodology: The sources of the papers were the following electronic databases: PubMed, Scopus, and Cochrane Library. A total of 20 out of 947 needed further studies, and this was encompassed in the present meta-analysis. It identified diagnostic accuracy, predictive performance, and potential biases. Results: AI models demonstrated an overall diagnostic accuracy of 82%, primarily leveraging artificial neural networks (ANNs) and convolutional neural networks (CNNs). These models have significantly improved the diagnostic precision for dental caries compared with traditional methods. Moreover, they have shown potential in detecting and managing conditions such as bone loss, malignant lesions, vertical root fractures, apical lesions, salivary gland disorders, and maxillofacial cysts, as well as in performing orthodontic assessments. However, the integration of AI systems into dentistry poses challenges, including potential data biases, cost implications, technical requirements, and ethical concerns such as patient data security and informed consent. AI models may also underperform when faced with limited or skewed datasets, thus underscoring the importance of robust training and validation procedures. Conclusions: AI has the potential to revolutionize dentistry by significantly improving diagnostic accuracy and treatment planning. However, before integrating this tool into clinical practice, a critical assessment of its advantages, disadvantages, and utility or ethical issues must be established. Future studies should aim to eradicate existing barriers and enhance the model's ease of understanding and challenges regarding expense and data protection, to ensure the effective utilization of AI in dental healthcare.