Artificial Intelligence in Dentistry

Abstract

In 1978, Richard Bellman defined artificial intelligence (AI) as the ‘mechanization of cognitive processes inherent to human thinking skills’. The first AI systems in medicine were developed in the 1950s when researchers at Jack Whitehead’s Technicon Corporation built a computer programme called the ‘MIT Programmed Autoanalyser’ to analyse the blood and urine samples. In the 1980s, machine learning algorithms were used in medical image analysis and drug discovery. The application of AI in dentistry is gaining popularity in pathology, dental radiology, caries detection, electronic record keeping and robotic assistance. Nguyen has classified the AI-based methods into four groups: statistical learning, artificial neural network (ANN)-based methods, genetic algorithm-based methods and hybrid AI methods. Statistical learning methods utilise the statistical principles and concepts to make predictions or decisions based on the data. These techniques involve fitting mathematical models to data and utilising statistical techniques for inference. Statistical learning methods can help analyse and model dental data, such as patient demographics, clinical measurements or treatment outcomes. Statistical learning techniques can also assist in predicting the risk of dental diseases, understanding factors affecting oral health, analysing treatment effectiveness and identifying patterns or associations in dental datasets.[1] ANNs are a type of machine learning algorithm that takes inspiration from the structure and functioning of the human brain. These algorithms utilise interconnected layers of artificial neurons to process and transmit the information. Neural networks can learn complex patterns and relationships in the data by adjusting the strengths of connections (weights) between the neurons. The examples of neural network-based methods include feedforward neural network, convolutional neural network (CNN) and recurrent neural network (RNN). In dental imaging, CNN can be employed for tasks such as tooth segmentation, dental anomaly detection or image-based diagnosis. RNN can be used in dental time series analysis, such as predicting orthodontic treatment progress or modelling temporomandibular joint disorders. Genetic algorithms are a type of optimisation technique that is inspired by the process of natural selection and genetics. They are used to search and optimise the solutions to complex problems. Genetic algorithms can be applied to various dental problems, such as treatment planning, optimisation of dental prosthetics or orthodontic treatment design. Genetic algorithms can be utilised for optimising the parameters in dental implant placement, designing optimal dental appliance structures or determining personalised treatment plans. Lee et al. proposed machine learning for the early detection of initial dental caries.[2] In their study, bitewing radiographic was used to train the U-shaped deep CNN (U-Net) model. Mohammad-Rahimi et al. in a systematic review reported that the accuracy of caries classification models was 71%–96% on intraoral images, 82%–99.2% on periapical radiographs, 87.6%–95.4% on bitewing radiographs, 68.0%–78.0% on near-infrared transillumination images, 88.7%–95.2% on optical coherence tomography images and 86.1%–96.1% on panoramic radiographs.[3] The differences in coverage area, image quality, lesion characteristics and dataset quantity may contribute to the observed variation in accuracy for caries detection between the imaging modalities. CHALLENGES Medical and dental data are not as available and accessible as other data, due to data protection concerns and organisational hurdles. Data are often locked within segregated, individualised and limitedly interoperable systems. Datasets lack structure and are often relatively small, at least when compared with other datasets in the AI realm. Data on each patient are complex, multi-dimensional and sensitive, with limited options for triangulating or validating it. Processing data, measuring and validating results is insufficiently replicable and robust in dental AI research. It is usually not possible to define a gold standard, and there is no agreement on how many experts are required to label a data point and how to merge different labels of such ‘fuzzy’ gold standards. The outcomes of AI in dentistry are often not readily applicable. The single information provided by most of today’s dental AI applications will only partially inform the required and complex decision-making in clinical care.

Ähnliche Arbeiten