Using AI to Detect Dental Caries

Artificial Intelligence (AI) is a branch of computer science that focuses on the creation of bodies of intellect in the form of software programs. Historically, AI has been known for its use in task-based systems, database analytics, and domain-specific knowledge, requiring manual fine-tuning by experts. However, machine learning (ML), a subset of AI, enables systems to learn intelligent tasks without the need for predefined boundaries or rules in data analysis. Deep learning (DL), another sub-branch of ML focuses on learning and composing patterns, and creating impactful systems, transferable to medicine and dentistry.

Today, medicine predominantly employs machine learning and deep learning techniques in various applications. Deep learning as a machine-learning process has been vital for studies relating to chronic oral disease for various ages of groups from teenagers to adults. A deep learning model identifies caries with high accuracy and newly graduated dentists should have noticed that. The results were then compared with expert-only caries detection methods. In the advancement in caries detection, the rise of the deep learning process known as ANN has taken over due to its artificial neural detection network. CNN, another deep learning software proposes non-surgical and surgical methods to impose caries detection measures. CNN serves as a computer-based method to reduce human error and promote the standardization of data. Google LLC takes all of the data that has been collected by systems and algorithms and picture projection, and analyzes it through a neural network, helping dentists make sense of data and its connections in caries detection.

Artificial intelligence models have proven effective in various applications in caries detection through the use of AI models such as convolutional neural networks (CNN), artificial neural networks (ANN), and Google LLC, used to take images and scans of patients caries development. By leveraging these AI tools, dentistry benefits from more efficient and precise diagnosis, ultimately leading to improved patient care and outcomes.

Traditional caries detection measures involve the usage of visual inspection, radiography, and forms of tactical sensation to guide professionals in clinical evaluations. Global disparities in health and oral care have indicated a need for cost-effective techniques to further bolster the prevalence of caries detection measures for the public, however. Although dentists’ ability to asses and diagnose caries is widely regarded as satisfactory, the standardization of the measurement has been a recent call to development to better asses in caries diagnosis. The field of dentistry has seen the development of techniques such as the intraoral camera (IOC) to combat the variability of dentists to provide aspects such as affordability, ease of operation techniques, and ease of storage as it utilizes digital image capabilities. Studies have proven that the usage of IOC offers a promising alternative to traditional methods of caries detection, reducing variability in results. Moreover, to better the process of standardizing dental caries detection, convolutional neural networks (CNNs) have been developed and have shown remarkable success in various medical applications. In dentistry, CNNs have been applied to detect carious lesions in radiographs and oral photographs, showcasing the potential for automated and accurate caries diagnosis through cutting-edge technology. The integration of advanced technologies not only enhances diagnostic capabilities but also holds future promise in improving real health outcomes.

In examining subjects' perceptions regarding AI software in dentistry, the study, "Patients’ perspectives on the use of artificial intelligence in dentistry: a regional survey," gathered insights from 265 participants regarding their knowledge and perceptions of this tool. A noteworthy finding was that 93.6% of patients assessed their digital technology knowledge as 'average' or 'above average,' while only 52.5% rated their knowledge of AI similarly. The study highlighted patients' concerns, with 37.7% expressing worries about the impact on workforce needs, 36.2% citing challenges to the dentist-patient fiduciary relationship, and 31.7% indicating concerns about increased dental care costs. On the positive side, 60.8% recognized 'improved diagnostic confidence' as a key advantage of AI in dentistry.

The evaluation metrics for caries detection, as shown in Tables 2 and 3, reveal the significant impact of image segmentation on the performance of classification algorithms. Results indicate that when segmentation was implemented, all key evaluation indexes saw notable improvements compared to cases without segmentation. Most notably, the AUC increased significantly from 0.731 to 0.831, and accuracy improved from 0.756 to 0.813, demonstrating the impressive ability of CNN algorithms to correctly classify 244 out of 300 test images for caries existence. Sensitivity and precision also showed marked improvement, further highlighting the effectiveness of segmentation. Furthermore, pre-processing of photographic images for tooth surface segmentation resulted in enhanced localization of carious lesions, further solidifying the benefits of segmentation in this realm.

Among all the images used, the CNN used in the study “Caries Detection on Intraoral Image Using Artificial Intelligence” accurately identified caries in 92.5% of cases and achieved an impressive 93.3% accuracy for cavitation detection. The diagnostic performance, however, varied depending on the type of caries lesion. Notably, caries-free surfaces were classified with the highest accuracy of 90.6%, followed by noncavitated lesions at 85.2% and cavitated lesions at 79.5%. Moreover, the study examined the effect of training data size on model performance. Surprisingly, even with just 25% of the available images, the model showed an overall agreement of approximately 80%. Increasing the training data to 50% resulted in a significant improvement in diagnostic performance, reaching around 90%. This highlights the importance of sufficient training data in achieving high accuracy in caries and cavitation classification.

Patients' opinions regarding the use of AI in dentistry have brought up several worrisome issues, including its impact on the workforce, potential obstacles to the dentist-patient relationship, and the potential for increased costs. To effectively address these concerns, it is crucial to prioritize data management, anonymization of personal information, and the establishment of regulatory frameworks. Fortunately, the study identified specific steps that can be taken to address safety concerns, notably through the implementation of the FDA's "Software as Medical Device" category. Additionally, it is crucial to establish clear accountability for any AI-related errors and advancements in transparency, especially concerning the use of ANN and CNN algorithms. Despite the challenges involved, the integration of AI, supported by emerging technologies from companies like Google LLC, holds tremendous potential for growth and progress in the field of dentistry.