Artificial intelligence-powered dentistry: Probing the potential, challenges, and ethicality of artificial intelligence in dentistry

Abstract

Introduction

Healthcare amelioration is exponential to technological advancement. In the recent era of automation, the consolidation of artificial intelligence (AI) in dentistry has rendered transformation in oral healthcare from a hardware-centric approach to a software-centric approach, leading to enhanced efficiency and improved educational and clinical outcomes.

Objectives

The aim of this narrative overview is to extend the succinct of the major events and innovations that led to the creation of modern-day AI and dentistry and the applicability of the former in dentistry. This article also prompts oral healthcare workers to endeavor a liable and optimal approach for effective incorporation of AI technology into their practice to promote oral health by exploring the potentials, constraints, and ethical considerations of AI in dentistry.

Methods

A comprehensive approach for searching the white and grey literature was carried out to collect and assess the data on AI, its use in dentistry, and the associated challenges and ethical concerns.

Results

AI in dentistry is still in its evolving phase with paramount applicabilities relevant to risk prediction, diagnosis, decision-making, prognosis, tailored treatment plans, patient management, and academia as well as the associated challenges and ethical concerns in its implementation.

Conclusion

The upsurging advancements in AI have resulted in transformations and promising outcomes across all domains of dentistry. In futurity, AI may be capable of executing a multitude of tasks in the domain of oral healthcare, at the level of or surpassing the ability of mankind. However, AI could be of significant benefit to oral health only if it is utilized under responsibility, ethicality and universality.

Introduction

Artificial intelligence

Artificial Intelligence (AI) can be construed as a computational system capable of emulating the cognitive abilities of the human mind, enabling it to acquire knowledge, engage in reasoning, and execute tasks or behaviors informed by acquired experiential data. ¹ AI is exemplified through the operation of machines, contrasting with the innate intelligence exhibited by humans. Fundamentally, AI encompasses algorithms that have been trained by computers with the objective of emulating human intelligence. Human intelligence denotes the innate cognitive capacity of humans that is biologically endowed, encompassing a range of abilities like perception, acquisition of knowledge, resolution of challenges, making choices, understanding language, and engaging in social exchanges.^2,3 AI principally operates via Learning—involves the collection of data and the establishment of principles for transforming the data into practical knowledge. These principles, referred to as algorithms, instruct the computer /machine on the sequential execution of a defined task. Reasoning—focuses on the identification of the optimal approach for attaining a specific outcome. Self-correction—ensures the delivery of the most precise outcomes by constant adjustment of algorithms. ⁴

In the contemporary era characterized by rapid digital advancements, AI has surfaced as a disruptive technology that has fundamentally altered numerous sectors and facets of our day-to-day existence. AI leverages the disciplines of life sciences, social sciences, healthcare, security, climate, and other technologies to simulate or imitate human intelligence using machines.^5–7 The common subdomains/disciplines of AI are:

• Machine learning (ML)—Teaching/training machines algorithms to make predictions/decisions using available data. It can be supervised, unsupervised, or semisupervised. Deep learning (DL) is a subfield of AI that seeks to replicate the cognitive processes of the human brain through the analysis and utilization of unstructured data for decision-making purposes, Natural language processing pertains to the capacity of AI to comprehend, interpret, and produce human language, facilitating machines in the comprehension and response to textual or verbal input and the artificial neural network (ANN) represents an artificial neuron, embodying a mathematical nonlinear model that draws inspiration from the human neuron, to address a particular task such as image classification (for instance, determining whether a radiographic image depicts a decayed tooth: yes or no) are some of the important learning models used in ML. ⁸ Robotics—a field within the realm of AI, which is dedicated to the utilization of programmed artificial agents known as robots. Expert systems—an AI computer system that mimics and learns from human experts’ decision-making skills. ⁹ Fuzzy logic—fuzzy logic embodies a computational methodology grounded in the concept of “degrees of truth” rather than the conventional binary logic prevalent in modern computing. The application of fuzzy logic is prevalent within the realm of medicine, particularly in addressing intricate challenges associated with decision-making processes.^10,11

AI, being an interdisciplinary field with various methodologies, is currently undergoing a significant transformation in numerous sectors, predominantly due to the progress made in ML and DL. At the core, all these AI tools are used to make human tasks easier and better. ¹² AI systems have a myriad of applications within the healthcare sector, offering the potential for quicker and more thorough decision-making assistance than what practitioners are capable of achieving independently. ¹³ AI systems have demonstrated a remarkable degree of precision in detecting tumors associated with breast cancer, retinal ailments, and more recently, in swiftly diagnosing pulmonary conditions related to corona virus disease of 2019 (COVID-19).^14,15

The worldwide AI market reached a valuation of USD 454.12 billion in the year 2022 and is projected to reach approximately USD 2575.16 billion by 2032, advancing at a 19% compound annual growth rate from 2022 to 2032. Globally, AI is dominated by the United States (36.84%) followed by Europe (24.97%) and Asia Pacific (24.93%). ¹⁶

Chronicle overview of AI

The origins of Modern AI can be traced back approximately seven decades, when the term “Artificial Intelligence” was first introduced during a workshop—Dartmouth Summer Research Project on Artificial Intelligence led by John McCarthy, Nathaniel Rochester, Marvin Minsky, and Claude Shannon at New Hampshire in 1956. ¹⁷ From the 1950s till date, the AI journey has witnessed several ups and downs influenced by numerous factors. ¹⁸ Figure 1 depicts the key advancements in AI since its inception.

Figure 1.

Chronicle overview of artificial intelligence (infographic).

Dentistry

Dentistry is the healthcare domain that deals with the prevention and management of oral diseases, encompassing conditions that affect the teeth, supporting structures/periodontium, and soft tissues of the mouth as well as the promotion of oral health. ¹⁹ It also involves addressing jaw deformities, dental misalignments, and congenital abnormalities like cleft lip and palate within its scope. Apart from general dental practice, dentistry comprises various specializations and subspecialties, such as orthodontics, pedodontics, periodontics, endodontics, prosthodontics, oral and maxillofacial surgery, oral and maxillofacial radiology, and public health dentistry. ²⁰

Good oral health is crucial to well-being and contributes significantly to the physical, psychological, social, and financial aspects of wellness. ²¹ World Health Organization (WHO) Global Health Status Report of 2022 projected that approximately 3.5 billion people globally are affected by oral diseases and most (3/4) of the affected individuals live in middle-income nations. ²² Dentists are now required to possess a deep understanding of emerging technologies and their impact on oral health and overall well-being to cope with the increasing burden of oral health problems globally. ²³

Chronicle overview of dentistry and dental innovations

Dentistry, tracing its origins to 7000 BC with the Indus Valley Civilization, stands as a longstanding medical profession. Dentistry, like other healthcare fields, has evolved through the ages. According to the American Dental Education Association, the Sumerians discussed tooth decay in their text in 5000 BC, followed by the Egyptian era in 2600 BC, with the well-known figure of Hesy Re, the first dentist. The Chinese practiced restorative dental procedures around 200 BC using silver fillings. The Greeks, Romans, and Arabs continued practicing dentistry culminating in the birth of modern dentistry in the 1700s with the popular figure of Pierre Fauchard, the Father of Modern Dentistry, followed by many advancements in the 1800s and 1900s.^24,25 The 20th and 21st centuries evidenced considerable innovations in dentistry, highlighted in Figure 2.

Figure 2.

Chronicle overview of dental innovations (infographic).

Coupling AI and dentistry

From the beginning of human civilization, technological advancements have played a significant role in alleviating the workload for individuals, thereby enhancing productivity and allowing individuals to explore other areas of alternative pursuits. Like other general-purpose technologies, for instance, electronics, automobiles, computers, and the internet, AI in its various forms (mentioned in the start) has transformed all sectors of human activities, spanning healthcare, education, manufacturing, finance, transportation, and digital media.^26–28 According to Forbes, the key areas of AI implementation in the healthcare industry include optimizing operational workflows, analyzing medical imaging data, incorporating robotic assistance in surgical interventions, and supporting healthcare professionals in their clinical decision-making processes. ²⁹ Dentistry places significant dependence on digital workflows, with an escalating integration of AI within the discipline. The utilization of AI in dentistry spans various areas including diagnosis and treatment strategizing, image interpretation, patient administration, prognostic analytics, and automation, elevating the quality of patient oral healthcare. ³⁰ Moreover, AI-driven virtual dental assistants can provide patients with tailored oral health recommendations and prompts for routine examinations, as well as guidance on proper hygiene practices. This empowers individuals to engage in proactive approaches to uphold their oral health at an optimal state. ³¹ World Dental Federation (Fédération Dentaire Internationale - FDI) recognizes AI as a pivotal technology within the dental field, particularly in relation to its potential impact on FDI's Vision 2030 initiative aimed at advancing comprehensive oral healthcare for global populations. ³² The market for AI in dentistry is projected to attain a valuation of USD 1.3 billion by 2026. ³³ The marketplace has the potential for segmentation based on various applications such as diagnosis, treatment planning, and patient management. It is anticipated that the diagnosis and imaging segments will dominate the market share.

Methodology

This narrative review explores and outlines the potential applications of AI in dentistry and the associated challenges and ethical considerations that arise in its implementation in this field. To present a wide overview of the topic, no strict criteria were applied like systematic reviews. To maintain its comprehensiveness, the white literature was explored via academic databases like PubMed, Scopus, Google Scholar, and IEEE Xplore while grey literature like government documents on AI, reports, online resources, and working papers on AI were searched via Google using keywords like “AI foundation and working,” “AI history,” “dentistry,” “history of dentistry,” “applications of Al in dentistry,” “AI in dental practice,” “AI implementation in dentistry,” “AI challenges in healthcare/dentistry,” and “AI and ethical issues in healthcare/dentistry.” The search strategy was further amplified and broadened using synonyms and Boolean operators. This review included literature from a diverse range of period, covering the foundational (historical) as well as the updated information relevant to the key themes of the topic. While, unauthentic literature and literature that did not encompass AI, its applications in dentistry, its ethical concerns, and associated challenges were excluded.

Results

The subsequent paragraphs provide a succinct overview of the AI application in various facets of dentistry aimed at enhancing the oral health of the population, and the dental industry as well as the associated challenges and ethical concerns (Figure 3).

Figure 3.

Applications of AI in dentistry (infographic).

ANN: artificial neural networks; AugI: augmented intelligence; DL: deep learning; FL: fuzzy logic; ML: machine learning.

AI in periodontology and implantology

AI has been utilized in the field of periodontology and dental implantology for disease risk assessment and prediction, diagnosis, prognosis, and treatment planning. AI systems are widely used in periodontal disease risk assessment and treatment planning.^34–37 An AI system developed by Chifor and colleagues aids in the automatic detection of normal tissues of the periodontium using ultrasound images. ³⁸ Similarly, Nguyen et al. fabricated a convolutional neural network (CNN)-based AI system for detecting cemento-enamel junction and measuring soft tissue height. AI algorithms have also been used in periodontal assessment to identify changes in soft tissues, bone density, alveolar bone loss, and classify the stages of periodontitis on radiographs. ³⁹ A DL model used in a study by Liu et al. detected gingivitis with accuracies between 77% and 94%. ⁴⁰ AI algorithms practiced in other studies accurately detected gingival inflammation plaque control efficacy.^41–44 Danks and coworkers used an AI system for automatic detection of anatomical boundaries and periodontal bone loss with improved accuracy. ⁴⁵ Kurt Bayrakdar and colleagues formulated a CNN model rooted in GoogleNet Inception v3 for the identification of loss of alveolar bone on orthopantomogram (OPG) images. Their model attained 90% accuracy in differentiation of alveolar bone loss. ⁴⁶ A hybrid fuzzy logic system used by Parewe and colleagues detected gingivitis and periodontitis with an accuracy of 82%. ⁴⁷ A DL model created by Chang and colleagues demonstrated superior efficacy in the identification and staging of periodontal disease from panoramic radiographic images and gingivitis using intraoral imaging. ⁴⁸ Thanathornwong and Suebnukarn developed an automatic regional CNN model for the detection of periodontally compromised teeth on OPG. ⁴⁹ Similarly, another CNN model indicated an accuracy of 79% in the identification of periodontally compromised teeth (posterior) using periapical radiographs. ⁵⁰ Other DL models automatically determine and classify loss of alveolar bone, using OPGs.^51,52

Nakano and colleagues formulated a multilayer perceptron model to predict oral halitosis based on salivary microbiota, achieving a precision of 0.96. Other AI (DL) systems using saliva and GCF detect the presence and severity of periodontitis in individuals. ⁵³

AI models mainly based on CNN have shown exceptional results in the detection of different problems relevant to dental implants like detection of dental implant type, guided implant surgeries, peri-implant bone loss, and implant fracture using different radiographic images.^36,54,55

CNN models utilizing cone beam computer tomography (CBCT) imaging have shown significant potential in detecting the morphology of alveolar bone at the edentulous areas as well as the surrounding anatomical structures like nasal fossa, maxillary sinus, and mandibular canal.^54,56 Additionally, CNN frameworks are being used in detecting and classifying implant fractures. ⁵⁷ An AI model (VGG-16) evaluated by Sukegawa et al. classified implants with the highest performance. ⁵⁸ AI model developed by Sakai et al. on CBCT images showed a high success rate for planning implant drilling during implant surgery (primary stability). ⁵⁹ Detection and classification of peri-implant bone loss have also been achieved accurately by CNN models using periapical radiographic images.^55,60 Additionally, ML algorithms are also being used in three-dimensional bioprinting for making tailored tissues, scaffolds, three-dimensional printed implants and preoperative surgical planning. AI tools also aid in accurately determining osteointegration, which portrays the success of dental implants. ⁶¹ Different augmented tools and software are also available for digital radiography, recording of periodontal phenotype, and clinical parameters like gingival margin, pocket depth, clinical attachment levels, tooth mobility, and furcation involvement.^62,63 Currently, Overjet, Pearl, VideaHealth, Perimetrics, and Dexis are the key commercial companies that have received food and drug administration (FDA) clearance for various AI-enabled softwares, software as a medical device (SaMD) used in dentistry. ⁶⁴

Apart from these AI systems, robots are also revolutionizing and augmenting dental implantology, providing virtual treatment planning, visualization, guided placement of the instrument, and haptic feedback. ⁶⁵ YOMI developed by Neocis is the first and only FDA—approved robotic system for implant surgery. ⁶⁶ Similarly, a team from the Rochester Institute of Technology has developed a prototype robot—voice-activated robotic dental assistant, which works on voice command to deliver sterilized instruments to the dentist. ⁶⁷

AI in restorative dentistry and endodontics

AI has revolutionized endodontics and restorative dentistry from the detection of dental caries to the identification of periapical lesions, roots and canals morphology, estimation of working length, root fractures, and treatment prognosis. ⁶⁸ Various studies have shown that AI models diagnosed dental caries with accuracy levels between 76% and 88.3%. ⁶⁹ A study by Devito and colleagues assessed an AI model made for the detection of proximal caries in bitewing radiographs, revealing that the AI system exhibited superior diagnostic performance compared to the most proficient human evaluator. ⁷⁰ Valizadeh and colleagues examined an AI algorithm designed for the purpose of diagnosing proximal caries based on periapical radiographs. This algorithm demonstrated a diagnostic accuracy of 97% for dentin caries, while its accuracy for diagnosing enamel caries was reported at only 60%. ⁷¹ In another study dental caries were identified using a CNN model on periapical images by Lee et al. The precision level for the identification of dental caries in premolars was 89.0%, molars was 88.0%, and premolars–molars was 82.0%. ⁷² AI models developed by Bayrakdar and coworkers for dental caries have shown superior performance compared to an oral-maxillofacial radiology resident with 2 years’ experience and a restorative dentistry resident with 3 years’ experience. ⁷³ Similarly, AI models (based on DL) designed by Cantu et al. for dental caries showed higher sensitivity (75%) as compared to dentists (36%). ⁷⁴ Another CNN model fabricated for dental caries detection on bitewing radiographic images by Srivastava and coworkers attained a notably higher sensitivity, that is, 81% in comparison to general dental practitioners, that is, up to 48%. ⁷⁵ Apart from caries detection, AI models can also aid in detecting surface loss from teeth. An ANN AI model used by Al Haidan detected tooth surface loss with more than 80% accuracy. ⁷⁶ Moreover, AI has the capability to identify the finishing lines during tooth preparation. This advancement can enhance the accuracy of dental restorations, resulting in superior adaptation and durability of the crown/bridge. ⁷⁷ Furthermore, various AL algorithms like fuzzy logic, ANN, and DL are used for shade and color matching.^78,79

A thorough knowledge of the root canal system is crucial for successful endodontic treatment outcomes and the usage of AI technologies holds promise in enhancing such diagnostic procedures. ⁸⁰ Several CNN architectures have been devised for the automated recognition, partitioning, and categorization of root canals depicted on plain and three-dimensional radiographs. The efficacy of these AI models has been proven to be comparable to or potentially surpassing that of dental professionals.^81–83 Findings from a study conducted by Yang and colleagues revealed that DL models showed elevated accuracy in predicting the presence of C-shaped canals using periapical and orthopantomogram (OPG) images. ⁸⁴ Another CNN model by Joe et al. demonstrated a notable accuracy of 95.1% in identifying C-shaped canals on OPG in lower molars. ⁸⁵

Periapical lesions, frequently arising as a result of pulpal pathologies, pose a significant challenge in terms of precise identification through conventional means. Nevertheless, the emergence of AI has introduced novel frameworks capable of aiding in this endeavor. ⁸⁶ A CNN model developed by Moidu et al. demonstrated favorable results in detecting and classifying periapical lesions, on periapical radiographic images. ⁸⁷ A study conducted by Li et al. on peri-apical lesions utilizing CNN on periapical radiographic images showed a high accuracy in the detection of such lesions. ⁸⁸ A CNN model developed by Pauwels and colleagues for the detection of periapical lesions outperformed human observers. ⁸⁹ Other studies based on AI models showed the same promising results in the automatic detection and classification of periapical pathologies.^90–92

Diagnosing root fractures (vertical and horizontal) is a difficult process that takes expertise. DL algorithms, specifically ANN architectures, have been designed to autonomously identify root fractures present in both two-dimensional and three-dimensional dental radiographs. Research conducted by Johari and colleagues identified vertical root fracture by employing a probabilistic neural network, demonstrating outstanding effectiveness by achieving an accuracy rate of 96.6% on cone-beam computed tomography (CBCT) as well as OPG. ⁹³ Similarly, CNNs have been shown by Fukuda et al. to be a potentially useful diagnostic element for vertical root fraction detection on OPG images. ⁹⁴

A good AI should have the potential to accurately predict the treatment outcome. In a case-based report by Campo et al. a reasoning paradigm was used for the prediction of treatment outcome of nonsurgical root therapy, with merits and demerits. This system which gathered data from areas such as performance, recall, and statistical probabilities, concluded whether one should go for retreatment or not. ⁹⁵

Another essential requirement for a successful endodontic treatment is the accurate detection of root canal apex. Techniques commonly used by endodontists are electronic apex locators along with radiography. ⁹⁶ It was analyzed that ANN can be utilized as a significant second opinion in locating the apical foramen on radiographs with 96% accuracy, thus enhancing the accuracy of determining the working length. ⁹⁷

Robotics have also transformed the field of endodontics. Apart from providing assistance, microrobots act as disinfectants to disrupt the oral biofilms within the root canal. These microrobot systems could also be used for the prevention of dental caries and peri-implant infections. ⁹⁸

While the AI models showcased in the research exhibit encouraging results, it is imperative to recognize the presence of specific constraints within the studies. To effectively tackle these constraints, consideration should be given to training and testing of AI models. Incorporating samples from multicenter and utilizing a variety of radiographic instruments is crucial for bolstering the resilience and generalizability of these AI models. ⁹⁹

AI in oral and maxillofacial surgery

AI algorithms can be used to help decisions in diagnosis, preoperative surgery planning, and outcome prediction in the domain of oral and maxillofacial surgery. ¹⁰⁰ Impacted third molar—which is the most prevalent oral surgery problem can now be detected with AI-based technologies as well as their complexity levels.^101,102 A deep CNN model was used in a study by Yoo et al. to assess the extraction difficulty of third molars on OPG. The success rate of this model in identifying the link with the ramus, angulation, depth and angulation was 82.03%, 90.23%, and 78.91%, respectively. ¹⁰³ Zhang et al. predicted facial swelling post third molar extraction, using an ANN model. The accuracy of this AI model was reported to be 98.0%. ¹⁰⁴

AI methods have shown promise in the detection and sectionalization of maxillofacial and dento-alveolar fractures, automated diagnosis of osteoarthritis in the jaw joint, measurement of the cortical thickness of the mandibular condylar head, and the diagnosis of condylar fractures on OPG and CBCT images.^105–107 Mandibular fractures can also be automatically detected and classified using AI CNN systems.^108,109 Soe et al. utilized DL for diagnosing fractures of the nasal bone. ¹¹⁰ Other AI models based on DL can aid in detecting skull fractures from CT scans and radiographic imaging.^111,112 Moreover, AI has also facilitated orthognathic surgeries which are performed in collaboration with orthodontists in many ways. ¹¹³ AI models have also demonstrated efficacious outcomes in the automated identification of temporomandibular joint-related disorders. ¹¹⁴

AI in prosthodontics

Prosthodontics is the field of dentistry that focuses on the diagnosis, treatment planning, preservation and rehabilitation of oral structures and their functions, and health of the patients with clinical problems associated with missing dentition and other maxillofacial tissue. ¹¹⁵ AI has significantly improved prosthodontics in recent years. These improvements include shade matching, computer-aided design/computer-aided manufacturing (CAD/CAM) restorations, implant surgery template creation, aesthetic dentistry, and debonding prediction. All elements associated with the designs of the prosthesis like designs of the occlusal surface for the crows, identification of the emerging profile in dental implants, and automated framework designs for removable partial dentures and complete dentures are all made possible by these AI technologies. ¹¹⁶

CNN has been employed for the categorization of dental arches and aiding in the production of removable partial dentures, providing a high degree of customization and comfort. Takahashi et al. conducted a systematic study to design an AI system that could categorize dental arches with accuracies of 99.5% and 99.9% for the upper jaw and lower jaw, respectively, along with assistance in manufacturing partial dentures by utilizing CNN. ¹¹⁷ the realm of fixed partial dentures, AI proves to be beneficial across various decision-making models, configurations, and results. The application of AI proves advantageous in shaping the occlusal morphology of the crown to align with the characteristics of the opposing dentition. Furthermore, AI facilitates the delineation of the emergence profile and the improvement of aesthetic aspects in fixed partial dentures. ¹¹⁸

Debonding of the crown/bridge is a prevalent issue faced by the patient can now be detected via AI tools. Yamaguchi et al. used a hybrid AI model based on DL and CNN for detecting the chances of debonding of CAD/CAM composite crowns on three-dimensional models with an accuracy of 98.5%. This AI system can be used in other sensitive areas as well such as die fracture. ¹¹⁹ Raith et al. designed an ANN model for determining cusp features on three-dimensional images, which showed an accuracy of 93.5% and aided in CAD/CAM designing. ¹²⁰ AI also assisted in the shade matching of various ceramic materials ⁷⁸ . Another study showed higher accuracy, higher marginal integrity, and reduced time of production for crowns designed via AI in comparison to digital and conventional wax-up. ¹²¹

Furthermore, robots can also be utilized for various tasks and procedures in prosthodontics. For instance, Otani et al. experimented with a robotic arm for tooth preparation on the maxillary central incisor. The robotic arm showed higher accuracy in terms of finish lines preparation. ¹²²

The concomitance of AI and intraoral scanners has transformed the field of prosthodontics by making it possible to forecast and build prostheses with remarkably accurate results in a matter of minutes. This has led to more consistent treatment outcomes and a significant reduction in the amount of laboratory time needed by dental technicians. In the future, the need for AI algorithms will increase with the advancement in CAD/CAM technology. ¹²³

AI in orthodontics and dentofacial orthopedic

A crucial part of successful orthodontic therapies is comprehensive/accurate diagnosis and treatment planning, which are difficult and subjective processes. Clinical decision assistance systems powered by AI aid doctors in reducing these issues. ¹²⁴ Cephalometric analysis (CA) is an important diagnostic tool in orthodontics. In comparison to manual CA, automated CA is more stable and repeatable, which is crucial for reliable CA. ¹²⁵ A study by Hwang et al. determined that automated CA can be authentic as an expert human reader. ¹²⁶ Other software tools that can construct cephalometric tracings on two-dimensional and three-dimensional radiographs using AI (CNN-based) were used by some researchers.^127,128 Moreover, AI has resulted in the improvement of the workflow of practices and reduction in the analysis time by up to 80 times in comparison to manual analysis. ¹²⁹ Similarly, the classification of skeletal patterns on radiographs and photos has been made easier by the development of AI models, which have shown great results with accuracy rates above 93% rates above 93%. ¹³⁰

One of the most challenging issues during orthodontic treatment is determining if extraction is necessary in a particular situation. Several AI tools have been developed in recent years to support therapeutic decision-making in orthodontics in recent years. With an accuracy score of 91%, ANN-based decision support systems have proven to be highly accurate in forecasting tooth extraction decisions especially as well as orthognathic surgery. ¹³¹ Findings from another CNN model study showed a success rate between 91.13% and 93.80%. ¹³² Li et al. reported a 94% accuracy in predicting extraction versus nonextraction. ¹³³ Another AI model based on ANN showed 80.0% accuracy in detecting orthodontic tooth extraction in children and adolescents. ¹³⁴

Accurate assessment of facial growth and development is essential for obtaining long-term outcomes in orthodontic treatment. ¹³⁵ AI algorithms have shown promising outcomes in the estimation of growth rate and development and its impact on orthodontic interventions. These encompass automated assessment of skeletal bone age, evaluation of maturation of cervical vertebrae, classification of skeletal patterns and analyzing the effects of orthognathic surgery on facial aesthetics.^130,136,137 AI models based on CNN can also be utilized for age estimation which is an important influencing factor in orthodontic therapy as well as forensic dentistry. ¹³⁸ AI has shown significant performance in detecting, diagnosing and classifying Temporomandibular Joint Osteoarthritis—a temporomandibular joint disease resulting in pain, dysfunction, and malocclusion. ¹³⁹

The application of AI in the domain of orthodontics has demonstrated significant advancements. Through ongoing AI education, effective regulation, and the resolution of privacy and ethical issues, the seamless incorporation of AI into orthodontic clinical settings will continue to thrive.

AI in oral and maxillofacial pathology

AI algorithms play a crucial role in detecting and estimating oral and maxillofacial abnormalities. AI models have been used to aid in the automatic identification of pathologies of maxillary sinuses for panoramic and CBCT imaging.^140,141 A study conducted by Keser et al. used an AI system based on DL for the identification of oral lichen planus through clinical images. ¹⁴² Similarly, an AI system based on ANN by Idrees et al. detected oral lichen planus with 94.62% accuracy. ¹⁴³ Campis and coworkers used a fuzzy logic system for the assessment of risk factors associated with oral candidiasis. ¹⁴⁴

Numerous studies have demonstrated the potential application of AI systems on two-dimensional and three-dimensional images for the identification of ameloblastomas, dentigerous cysts, odontogenic keratocysts, radicular cysts, follicular cysts, and diverse oral tumors with higher accuracy.^145–147

Postoperative infections after orthognathic surgery have been predicted precisely via AI tools with higher success rates.^113,140 Additionally, AI has exhibited its worth in the realm of oral cancer detection. Models based on CNN have shown the ability to autonomously identify oral cancer on both photograph and confocal laser microscopic representations.^148,149 The AI model based on DL, by Ariji and coworkers, analysis of oral cancers, and possible lymph node metastases, outperformed the radiologist. ¹⁵⁰ A CNN model based on photographs was developed by Warin and coworkers for an automatic detection and classification model for oral cancer. ¹⁵¹

AI systems present a hopeful outlook in enhancing the diagnostic abilities of oral health professionals and enabling the timely identification of oral pathologies, ultimately resulting in enhanced prognosis.

AI in public health dentistry

Prevention of oral diseases and promotion of oral health are the key domains of public health dentistry. AI enhances the potential of public health dentistry by potentiating preventive measures and promoting oral health, enhancing the oral health services and surveillances by providing aids in early screening of oral conditions, maintenance of electronic health records, and patients’ schedule management and surveillance of oral health issues.^152,153 For instance, the detection and prevention of early childhood caries by AICaries—a smartphone application ¹⁵⁴ or AI-driven optimization approaches for early diagnosis, and improvement of obstructive sleep apnea—a prominent dental health issue, via continuous positive airway pressure. ¹⁵⁵

Teledentistry—a domain of public health dentistry that provides oral care and instructions via televisions, phones, computer messaging applications, and social media, allowing far-off diagnostic and treatment recommendations. AI-enabled teledentistry has a wide range of applications such as screening, prioritizing, diagnosing, monitoring, and providing feedback. This technological advancement is witnessing substantial expansion and holds the capability to substitute conventional in-person consultations under specific circumstances. ¹⁵⁶ AI-driven systems can assist in the remote diagnosis of oral conditions through the analysis of images, radiographs, and patient information, thus offering dentists crucial insights for treatment strategizing. ¹⁵⁷ For instance, MeMoSa is a mobile screening app that uses photographs to detect oral lesions with more than 80% accuracy. ¹⁵⁸ Furthermore, AI has the capability to facilitate remote monitoring of treatment advancements, as well as provide patients with feedback. ¹⁵⁹ AI-fueled chatbots and digital assistants enable remote consultations, where patients can inquire, exchange images, and obtain initial guidance, personalized treatment and tailored education, without the need to physically visit the healthcare facility. ³¹ Oral squamous cell carcinoma, lichen planus, and leukoplakia can now be easily assessed remotely from intraoral images. ¹⁶⁰

As technological advancements progress, AI-assisted teledentistry holds the capacity to transform the field of dental care, offering streamlined and successful dental services from a distance, diminishing obstacles to accessibility, and enhancing general oral health results for individuals.

AI in dental academia

Dental education involves an integration of theoretical, preclinical, and clinical education, covering both didactic and clinical skills training. The swift implementation of AI and other transformative digital technologies has introduced the possibility of AI-fueled educational platforms to improve the educational journey of contemporary dental students in both theoretical knowledge and practical skills. ¹⁶¹

In AI-enhanced learning settings, teachers may function effectively and support their professional development. The integration of AI technologies aids in promoting human interaction and collaboration among students, resulting in enhanced academic outcomes. ¹⁶² There is a growing curiosity surrounding the capabilities of AI chatbots and advanced language models like OpenAI's ChatGPT and Google's Gemini, particularly in the realm of dental education. Furthermore, the integration of these technologies may necessitate adjustments in the approach to composing essays, dissertations, or scientific papers. ¹⁶³

Incorporating AI-driven technologies like virtual reality, augmented reality, mixed reality, and metaverse into dental simulations and practical exercises in laboratory and preclinical education amplifies the efficacy of simulations for dental students.^164,165 For instance, SIMROID serves as a simulation system designed for training in a wide range of treatment methods and interpersonal communication abilities through the utilization of an exceptionally lifelike robotic patient. Apart from recording and reviewing the training operation, it has the capacity to provide human-like expressions. ¹⁶⁶ ROBOTUTOR is another robotic system that can be used for teaching purposes as well as oral health education. ¹⁶⁷ A study carried out by Schwendicke et al. sought to establish a fundamental curriculum regarding AI for both undergraduate and postgraduate dental education programs to cope with the oral health problems of this age of automation in an effective and automated way. ¹⁶⁸

The integration of AI-assisted tools and simulation technologies in educational practices is anticipated to enrich the educational experience and foster cognitive development, self-awareness, and self-assurance among dental students. As a result, dental practitioners equipped with modern skills tailored to the demands of the current era, incorporating AI and other cutting-edge technologies are poised to supplant conventional dental professionals in the imminent time frame.

Discussions

AI is a rapidly expanding technology encompassing every domain of life from life sciences to physical sciences to earth sciences, directly or indirectly. The amalgamation of humans and AI is probing the evolution of galaxies, exploring the depth of oceans, and unwinding the structure of proteins, drug designs, and genetics. ¹⁶⁹ Based on current observation, AI in the dental industry is accompanied by the following scenarios.

Competencies of AI in dentistry

AI has accomplished substantial advancements in the dental industry, specifically in the areas of image processing and analysis. One of the remarkable achievements of AI in dentistry is the prompt detection of numerous oral disorders like dental caries, oral carcinomas, and periodontal disease via DL algorithms. Such AI-powered agents not only help in timely disease identification but also aid in the reduction of discrepancies in their detection. Apart from early identification of disorders, AI is also proficient in analyzing large data sets (e.g. electronic health records and clinical decision support systems), which aid in detecting the progression and pattern of patients’ dental and overall health status. This facilitates tailored treatment planning resulting in better clinical prognosis. ¹⁷⁰

Incompetencies of AI in dentistry

Every human invention is accompanied by imperfection, and the same goes for AI. In dentistry, AI presents several shortcomings including a lack decision in complicated clinical scenarios. For instance, AI systems may not be capable of considering the psychological/empathy component of patient history, and variations in clinical presentation, unlike a skilled clinician in devising a treatment plan. Another drawback of AI is its high cost, which makes it impractical to be used in low-resource dental settings. The “black-box” nature of AI systems is another major flaw that makes it challenging for the dentist to fully comprehend how particular decisions are generated.^171,172

Practical implication of AI

Dental organizations

1. Patients’ management: AI helps to streamline patient management, including dental and medical history, and treatment records.

2. Automating the administrative tasks: AI reduces the labor and resources required for the smooth running of the administrative activities of dental organizations. These automation activities span from AI-powered chatbots (to facilitate patients by answering their queries) to automated billing and scheduling, resulting in enhanced operational efficiency and reduced cost. ¹⁷³

3. Forecast analytics: AI-powered analysis of patients’ perception, practice performance as well as the latest innovative trends in the dental industry may help the dental organization to timely revise and adapt their policies to stand firmly in the era of AI. ¹⁷⁴

Patient care

1. Better access to dental care: by surpassing the geographical barriers, AI-powered teledentistry can improve dental care via teleconsultation and telediagnosis.

2. Enhanced diagnostic accuracy: AI-backed diagnostic software not only aids in disease risk prediction but also facilitates early diagnosis of dental problems, improving the prognosis.

3. Individualized treatment plan: AI facilitates tailored treatment plans by taking into consideration individual's electronic health records including genomics. This results not only in better prognosis but also reduces the cost of treatments. ⁶⁹

Dental staff

• AI assists the dental workforce in a variety of ways. Following are the principal corners where AI has influential impacts.

1. Efficient routine tasks: AI helps the dental team to handle various clinical and nonclinical chores effectively through AI-powered tools which not only help in administrative tasks, prevention, diagnosis, treatment planning, and prognosis but also save time, cost, and labor (for instance, SaMD).

2. Enhanced education and training: AI-powered augmented technologies assist the dental workforce in training and education of the latest techniques and procedures, through the provision of virtual surroundings and scenarios.^164,175

Challenges associated with AI implementation in dentistry

According to Rick Stevens, Professor of Computer Science at the University of Chicago “whoever leads the world in AI for science will lead the world in scientific discovery.” ¹⁷⁶ But this lead comes at the cost of challenges, responsibilities, and regulations to integrate human-centered amicable AI into various sectors of society with resultant universal favorable outcomes. The coming sections will present an overview of such challenges and ethical considerations of AI in the domain of dentistry.

Ethical and privacy-related concerns

The incorporation of AI within the healthcare sector introduces a diverse array of challenges pertaining to privacy and ethics. These challenges encompass concerns related to safeguarding patient data, defining the ethical limits of technological advancements, and evaluating the tangible effects of technology on healthcare professionals as well as individuals seeking medical treatments. ¹⁷⁷

In order to fully harness the capabilities of AI in the healthcare sector, it is imperative to make strategic choices that carefully weigh various conflicting concerns and principles. These considerations encompass matters such as privacy, accountability, intellectual property rights, and the promotion of transparency.

Adoption and acceptation concerns

Healthcare professionals, such as physicians, nurses, and assistant nurses, exhibit an insufficient emphasis on digital technology, potentially impeding the integration of AI in the healthcare sector. Furthermore, the general population frequently exhibits a deficiency in comprehending the potential advantages and constraints of AI within the healthcare sector, resulting in impractical expectations that may impede physicians from embracing these technological advancements. ¹⁷⁸ Dentists must develop particular skills related to the utilization of AI to ensure the safe and efficient application of AI in the treatment of dental patients.

Biasness and partiality concerns

AI bias arises from a variety of factors, including incomplete data, biased sampling, biased model training, or other elements that can lead to unequal treatment, misdiagnosis, or underdiagnosis of specific demographic groups. AI systems acquire knowledge and understanding from the data on which they are trained, thereby potentially adopting the biases present in that data. For instance, AI systems developed using data primarily sourced from academic medical centers may exhibit reduced awareness and consequently provide less effective treatment for patient populations that do not typically seek care at such facilities. ¹⁷⁹ Continuous surveillance is imperative for the identification of such complications within the AI system.

High-cost concerns

Limitations in terms of finances and resources prevent AI from being widely used in dentistry, especially in resource-limited setups. Acquiring AI systems, hardware, and software licenses can be difficult for dentistry offices, especially smaller ones, because of restricted resources and funding. Challenges may arise during the shift from traditional computer structures to AI architectures, as the introduction of graphic processors, field-programmable gate arrays, and specialized AI chips necessitates sophisticated computing and storage mechanisms to support AI functionalities. ¹⁸⁰ Nevertheless, the implementation of such infrastructures and storage solutions may pose financial burdens on healthcare institutions.

Interoperability concerns

The significance of interoperability within the healthcare sector cannot be overstated as it plays a fundamental role in facilitating the provision of effective, patient-focused, and top-notch healthcare provisions. ¹⁸¹ In order to guarantee the effective execution and incorporation of AI in the oral healthcare sector, it is imperative for oral healthcare practitioners and AI specialists to collaborate. This partnership has the potential to pinpoint possible obstacles, develop customized remedies, and enhance diagnostic precision through the utilization of large datasets while reducing the opacity inherent in AI systems. ¹⁷⁰

While the utilization of AI in oral healthcare presents notable obstacles, these challenges can be effectively tackled by implementing stringent data security protocols, establishing ethical frameworks, and fostering collaborations to surmount technological and clinical impediments, thereby unleashing the complete potential of AI in enhancing the quality of oral healthcare services.

Ethical considerations

New technologies can bring about advantages for our society, but they can also bring about risks and hazards that need to be carefully considered and handled. AI possesses the capacity to transform how we engage in our daily lives and professional endeavors, and it is presently exerting a significant influence across various sectors. ⁵ Nevertheless, like any technological advancement, AI gives rise to moral and societal apprehensions, encompassing concerns such as privacy infringements, employment displacement, and partiality in the process of decision-making. It is imperative for organizations, policymakers, and governments at large to thoroughly ponder over these concerns and foster ethical AI protocols as the discipline progresses and reaches a state of maturity. Consequently, the current course of action necessitates the exploration of appropriate regulatory frameworks for AI, its potential applications for societal benefit, and the ethical considerations surrounding its utilization.^182,183 AI is analogous to a double-edged sword, can be wielded for the advancement of humanity or conversely, to its detriment. AI technologies, such as large language models, are being swiftly implemented, at times lacking a comprehensive apprehension of their potential outcomes, which could potentially have positive or negative impacts on end-users, encompassing healthcare providers and individuals seeking medical care. In the context of utilizing health-related data, AI systems might be privy to confidential personal data, underscoring the requirement for strong legal and regulatory structures to protect privacy, security, and authenticity. Every segment of society must actively participate in the transition to responsible AI. ¹⁸⁴ To protect and advance a balanced and human-centered use of AI in healthcare, the WHO's Working Group on Regulatory Considerations (WG-RC) drafted some regulations for AI in healthcare Table 1. ¹⁸⁵

Table 1.

WHO working group on regulatory considerations (WC-RC)

S. no	Regulation	Description
1	Transparency and documentation	Complete information about the AI product to ensure trust among developers, manufacturers, and end-users.
2	Risk management	All risks associated with AI health systems and devices, be extensively discussed.
3	Intended usage and data validation	It must be made clear to promote regulation and ensure safety.
4	Data standardization	Thoroughly test AI systems before release to make sure biases and faults are not amplified.
5	Confidentiality and data security	Accentuation on comprehending the boundaries of jurisdiction and the need for consent.
6	Collaboration	Promoting cooperation among government agencies, industry representatives, developers, patients, healthcare providers, and regulatory authorities may enhance the security and quality of an AI system.

AI: artificial intelligence; WC-RC: working group on regulatory considerations.

The goal of WHO guidelines is to provide a framework of fundamental principles for governments and regulatory bodies to use when creating new AI guidelines or modifying those that already exist at the national or regional level in the healthcare industry. ¹⁸⁶

Since many of the hazards associated with AI are intrinsically global in scope, multinational collaboration is the most effective way to solve them, involving governments, organizations, civil society, and academia like the organisation for economic co-operation and development (OECD)'s AI Principles, the US Executive Order on Safe, Secure and Trustworthy AI, Canada's C-27 AI and Data Act, the AI4People's 7 AI Global Frameworks (2020), UK's AI Safety Summit—the Bletchley Declaration (2023), Convention on AI, Human Rights, Democracy and the Rule of Law by the Council of Europe, the United Nations (UN) Global Digital Compact, the G7 AI Principles and Code of Conduct (AIP&CoC), the EU AI Act (2024), and United Nations Educational, Scientific and Cultural Organization (UNESCO)’s Recommendation on the Ethics of AI (2021).^187,188

Similarly, GlobalPolicy.AI has unified the work of eight global organizations (the Council of Europe, the European Commission, the European Union Agency for Fundamental Rights/FRA, the Inter-American Development Bank, the OECD and OECD Artificial Intelligence Policy Observatory/OECD.AI, the UN, the UNESCO, and the World Bank Group) on a responsible AI synchronized with the human rights and democracy. GlobalPolicy.AI enables member organizations to exchange practical AI resources and best practices that can assist in achieving the sustainable development goals in a variety of fields, including agriculture, sustainable cities, transportation, health, and education by forming alliances with important participants from all stakeholder groups who have similar principles and aspirations for the advancement of reliable, human-centered AI in the future. ¹⁸⁹

AI health systems and technologies that are founded on ethical principles hold the promise of progressing human rights concerns through enhanced availability of diagnosis, individualized treatment options, and educational resources. ¹⁷⁷ Health authorities (like the FDA in the United States) have released criteria for evaluating AI-based medical devices. In 2019, a regulatory framework “Software as a Medical Device (SaMD)” was introduced by the FDA for the modification of AI/ML-based devices/software. As per the 2023 report, the FDA has listed 691 SaMD, the majority of which are related to radiology (531 or 77%) ¹⁹⁰ and only 2.8% are related to dentistry. ¹⁷⁵ Currently, VideaHealth, Overjet, Pearl, and Dexis are the prominent FDA-approved players in AI SaMD in dentistry. ⁶⁴

In addition to rigorous regulatory protocols, educational and instructional initiatives aimed at healthcare practitioners necessitate the integration of comprehensive discourse surrounding AI, encompassing its benefits, drawbacks, and strategies for risk mitigation. ¹⁶² According to the white paper by the American Dental Association regarding AI in the field of dentistry, the advancement of AI and augmented intelligence is steadily evolving, but it is crucial to keep holding the clinical judgments for dentists. The report further stressed adherence to ethical regulatory frameworks, safeguarding of data privacy and a conscientious deployment of AI and augmented technologies for the improvement of oral health. ¹⁹¹

Moreover, dental practitioners are advised to conduct a rigorous assessment of the certification status of AI products prior to their utilization in patient care. The standardization initiatives in this domain of AI applications exhibit a higher level of development, with numerous bodies such as the FDI offering established or in-progress standards and advisory materials. ³²

In brief, the ethical implementation of AI in the field of dentistry necessitates adherence to certain guidelines; specifically, AI systems must undergo approval by duly appointed regulatory bodies prior to their deployment, dental professionals should receive proper education and training for instance federated learning on the training and utilization of AI, and should consistently assess the efficacy of AI-driven systems within their patient cohort guaranteeing the well-being of patients and the security of their information.^192,193

Future prospects

Based on the current propitious amalgamation of AI and dentistry, the future of AI in dentistry is very up-and-coming. The current risk prediction assessment, diagnostics and prognostics will reach new heights of accuracy and precision soon. New drugs discoveries and methods of their delivery will pave the way for personalized medicine, powered by AI. Advancements in extended reality will make the use of augmented and mixed realities easily accessible and more prevalent, benefiting both the clinical domain as well as the academia in dentistry. Dental robots powered by AI are anticipated to become more daedal with multipurpose potentials and higher accuracies. Similarly, nanorobots will open new channels for local drug delivery and disease diagnosis. Regardless of the rise of AI-powered dental robots in carrying out recurrent tasks, it is doubtful that dentists will be completely substituted by robots in the anticipated future. ¹⁹⁴

Limitations

Although this overview projects a paradigm of AI applications in dentistry and associated challenges and ethical concerns but it is accompanied by the following limitations.

1. AI and its applications in dentistry are swiftly progressing, so this review may not represent the most recent innovations and associated concerns that emerged during or after its publication.

2. This review contains information from both white literature and grey literature, which may pose a discrepancy in the reliability of the grey literature in comparison to the former.

3. The ethical considerations addressed in this review gave an overview of such issues in the context of dentistry/healthcare. Since the ethical insights from other domains like social sciences may not be explored in depth, limiting the applicability of some ethical considerations.

4. Due to the narrative nature, this overview spotlights the paramount challenges associated with AI implementation in dentistry, but may perhaps not provide in-depth information about the strategies to overcome barriers in effective implementation of AI in diversified settings.

Conclusion

The coalescence of AI and dentistry holds significant promise for transforming the oral healthcare landscape. The multifarious domains of AI are continuously improving the accuracy of disease risk prediction, diagnosis, effectiveness of treatment, and overall oral care. By enhancing collaboration among dental professionals and experts in the field of AI, as well as incorporating specialized training focused on AI into dental courses, a proficient workforce can be cultivated with the ability to efficiently utilize AI for enhancing oral healthcare, all within the framework of universal ethical guidelines.