Since the origin of science, much research has been taking place to solve the human brain complexity. Human brain deals with neurons which are interconnected with each other, and they will help for signal transmission of the complete body. Scientists had studied a lot to design and develop a model which mimics or imitates human brain, but it has remained as a big puzzle to solve. Even though it is a big task for scientists, they have put a constant effort for so many years for the evolution of “artificial intelligence (AI).”
From the past few decades, applications related to AI have gained various advantages as well as curse. However, from the past few years, AI has gained amazing achievements in various fields. The main advantages of AI are that it is based on natural language model and has become most convincing where readers can't even distinguish if it is a human written or system written. Another advantage is face recognition system. Hence, based on various exciting advantages of AI-based technologies, various fields had gained many benefits including health care and dentistry is particular among them [Figure 1].
It is a process of assembling artificial neurons and these layers of artificial neurons are connected by various mathematical operations and a network is engineered which is capable of dealing with various specific tasks such as image classification. The AI can be involved by using the above process in locating canals during randomized control trial and radiographic image which shows damaged or decayed tooth.
MATERIALS AND METHODS
Clinical decision support system
In the past few years, health problems and hospital visits are increasing tremendously which resulted in continuous increase in patient data. Thereby clinical decision support systems (CDSSs) are becoming one of the vital tools for health-care providers and hospital management. These are the computerized programs which use past clinical knowledge in analyzing the data of the patients and assist health professionals including dental health professionals in making clinical decisions such as diagnosis or treatment of the disease.
This can be explained by a simple example such as a patient visiting a dentist with a compliant of toothache. In such cases, CDSS can be helpful in decision-making of dental health professionals by carrying out a patient risk assessment such as sugar intake and fluoride exposure and by suggesting a treatment plan. Thereby CDSS can provide value addition to both health professionals and patients by providing faster and systematic health outcomes.
Implementation of artificial intelligence in dentistry
As AI through CDSS started increasing its role with the availability of more patient data on a dynamic basis, its role in dentistry has been becoming more prominent in the recent times. Its role has been becoming more in diagnosis, patient management, orthodontics, and periodontics.
It can be inferred from many instances that diagnosis is the key for successful treatment of any disease. Artificial neural networks (ANNs) can be more helpful for the diagnosis of the health condition or dental condition, thus assisting dental personnel in timely treatment based on the risk assessment of the patient. This faster diagnosis is enabled by analysis of the reoccurrence of earlier dental condition and by minimizing human errors. Many studies had highlighted that the usage of ANN had been highly specific and thereby revealing the prominence of AI in dentistry diagnosis including complicated oral diseases. For prompt diagnosis of any oral disease, the patient data for the ANN are made available through digital Intraoral Periapical Radiograph (IOPA) X-ray, scans, etc. This is very prominent in identification of risk group who are more susceptible to oral cancers. The results of many studies inferred that genetic algorithms in ANN are very helpful in prediction of erupted canine sizes or premolars. It can also assist in prediction of tooth surface loss.
The evolution of AI paved way for virtual assistants to perform several tasks with greater precision. These virtual assistants can be used right from booking of patient's appointment, collection and analyzing patient's dental history, insurance management and also can assist the dental health professional in proper treatment planning based on the diagnosis. It can raise an alarm to the dental professional on the patient's health history and habits such as smoking.
AI enables in creation of a virtual database to assist the professionals in treatment and also helps in follow-up and emergency.
Customized orthodontic treatment options are considered the recent evolutions enabled by AI. AI is used in several phases such as diagnosis, treatment, follow-up, and monitoring. Further, with the usage of three-dimensional scans, the assessment of dental abnormalities has become more feasible and this enabled customization of treatment. It works on a data algorithm which decides on the teeth which should be moved and even decides on the pressure which should be pragmatic and provide the pressure points. This makes sure that precise treatment is provided with less time for treatment.
There are two clinical types of periodontal disease, which are aggressive and chronic types. Due to the complexity of the pathogenesis, there is a lack of evidence related to tests which can be discriminated between these two types of disease. This led to the usage of ANN by analyzing immunologic parameters to precise identification between types of the periodontal disease. Many studies concluded that AI is useful and can be deployed for accurate and error-free diagnosis of aggressive and chronic types of periodontal disease and also for the successful treatment of periodontically compromised teeth.
Cone-beam computed tomography (CT) is normally used in optimization of clinical outcomes related to endodontic therapy. However, it poses a challenge as it cannot be used systematically due to its higher dosage of radiation in comparison with other conventional radiographs. Deep learning algorithms are used for classification of morphology. However, it is also observed that though the algorithms and convolutional neural network (CNN) are very accurate, there are some limitations for implementation of AI in endodontics. These are manual segmentation of images which consume a lot of time and sizes of the images and specific accuracy related to the image focus.
Anthropological analysis and calculations, esthetics, and facial measurements are the factors which have very critical importance for the dentist to provide prosthesis as per requirement. The usage of technology in order to fit the prosthesis perfectly is another milestone achieved by AI in dentistry. Further, the systems based on computer-aided design/computer-aided manufacture helped in achieving high precision and finished dental restorations, and they are also used in design inlays and outlays. With the extensive benefits of AI implementation, it has become the most preferred option and replaced the conventional techniques of casting the prosthesis and also decreased the time required with less errors.
DETECTION OF ORAL CANCER
Early diagnosis is most important and lifesaving, especially in oral cancer patients. It is known that during the late stage of this disease, it has a very poor prognosis. Cytology, CT, and fluorescent images are used for data collection and can be used as learning tools for AI in diagnosis of the oral cancer and to the most precise extent. There are many researchers conducted on the same, and all these researches concluded that the application of AI-based learning models for detection of oral cancer is more effective and accurate. Many of these research results highlighted the early diagnosis in oral cancer cases which may occur from subsides of the oral cavity, by using models of AI. Results of the researches also highlighted that the efficacy and accuracy of the AI is 80%–85% when compared to conventional cytology. Further, the deployment of ANN-based models also showed betterment in malignant detection accuracy which moves to 93%. The sensitivity is at 95% when used regression-based algorithms in comparison to traditional algorithms.
The data collection methods such as brush biopsy are very minimal invasive which provides comfort to the patients. When compared to conventional methods, the usage of these algorithms also improved the quality of the diagnosis and even can be done through images taken by smartphones through concepts developed based on AI. White light imaging and autofluorescence are added to the images taken by smartphones in the above concept. These images will be stacked to the algorithms in order to detect or recognize any kind of oral malignancy. This strategy of smartphone-based image stacking made detection of oral cancer more convenient and accurate. However, these methods of oral cancer detection are at very nascent stages, and more studies and researches are needed to have proper validation of the technique.
Principal component analysis (PCA) which is a method based on computing on principal components of data is also proposed for diagnosis of oral cancer as this method is also minimally invasive. However, many research studies pointed out that ANN-based models are better performed compared to PCA. There are evidences from other studies that AI-based models are very efficient in detection of neck-and-head cancers.
PREDICTING THE OCCURRENCE OF ORAL CANCER
At present, there are many advanced treatment options which are available for the treatment of oral cancer. However, the major concern with the oral cancer is its higher rate of reoccurrence. The oral malignancy and its growth depend upon the stage of the cancer and any lack of evidence on prediction of the stage of the disease will lead to inefficient treatment plan. There are several prognostic biomarkers which are proposed in ongoing periods. However, these biomarkers cannot be reproduced in the current disease staging system. Several conventional statistical methods are being used for predicting the oral cancer; however, given the complexity of the data involved in oral carcinoma, it is found not suitable for conditions like oral cancer. Hence, for such complex datasets, AI-based prediction models are supposed to provide precise outcomes.
Several studies are conducted on the above-proposed models, such as research conducted by Alabi et al. in Brazil where it compared four machine learning algorithms for prediction of the reoccurrence of oral tongue squamous cell carcinoma. The results of this study concluded the improved accuracy in prediction of the stage system of the oral cancer and reoccurrence of the disease.
These machine learning algorithms are developed on basis of naïve Bayes, boosted decision tree (BDT), support vector machine, and decision forest. In all the above algorithms, the results of the studies observed that the BDT algorithm provided more accurate results. These models used several factors such as death, cancer occurrence, survival rate, and metastasis.
These research studies pointed out that the decision tree was more easily interpretable and accurate when compared to traditional logistic regression. Studies are also conducted to compare fuzzy regression model (which explains the relationship between response and explanatory variables), neural network prediction model, and the clinician opinion for prediction of the oral cancer. These studies concluded that the AI-based neural network and fuzzy regression models performed well in prediction of the disease compared to the human opinion.
The results showed that the accuracy of prediction of the disease is almost 90%–96% when the above algorithms are used.
However, as the datasets are small for these studies, extensive research and further studies are required for proper validation of the results. However, the evidence clearly suggest that AI-based machine learning algorithms are very effective compared to the conventional statistical methods.
Challenges faced by artificial intelligence implementation in dentistry
Future of the AI in dentistry mainly depends on addressing the key shortcomings in its use such as complexity involved in the system or mechanism, expensive setup, lack of proper training in the models, and data snooping bias. Further, the outcomes of the models are not sometimes readily applicable in the dentistry. There are several foremost challenges which are to be addressed for successful implementation of AI in health care. The clinical data sharing and management are very critical in successful implementation of AI systems. The AI systems and algorithms need personal data of the patients for validation and improvement of the system and often they share these data among different institutions also across the national boundaries. Hence, the patient data sharing and management have to be strictly followed as per laid down guidelines to ensure that the privacy of the individual is not affected. Further, the systems must require to adapt to protect the patient confidentiality and privacy in order to integrate AI and clinical operations. The personal data have to be anonymized before any sharing and broader distribution.
Even with these precautions and checks in place, the health-care community is still skeptical about the protection of the data in sharing. Hence, this challenge is to be addressed properly with transparency on the precautions taken to ensure successful implementation of AI in health care.
Further, the AI systems also have safety issues and hence there should be mechanisms in place in order to control the quality of the algorithms and systems used in AI. Another challenge in implementation of AI is ambiguous accountability in its use. Many times, there is no clarity on the accountability associated with any unintentional consequences faced by patients as a result of any adverse event in the AI system and the algorithm usage. Substitution of humans with autonomous agents gives scope to various questions regarding the legal and ethical order, and these issues pose challenge to the current legal system.
Transparency of the algorithms is another substantial challenge faced by AI in its implementation in health care and dentistry. The predictions heavily rely on the accurate annotations and labeling of the datasets used in the training of the algorithm. Hence, any poorly labeled data will cause distracted results. Further, the clinic-labeled datasets may itself be of inconsistent quality and thereby having significant effect on the efficacy of the AI systems.
Health-care authorities also should fully understand the decisions and predictions of the AI system in order to defend the decisions. Hence, interpretability of these AI systems and consequences has always been a challenge in the implementation of the technology. Hence, major advances are also required to ensure transparency and increase interpretability of the results and outcomes.
AI is no longer a myth but the future in dentistry. With its growing application in the recent years, it will be the most promising tool in the future in diagnosis and treatment of the dental problems. One cannot expect it will completely replace the role of dentists as the dental practice is not only about the diagnosis of the disease but also includes correlation of the findings and provides proper treatment to the patients.
However, there should be clear understanding of the concepts and models of AI in order to have full benefit of the technology. Dentists and clinicians also should ensure collection and providing authentic data into their database in order to have accurate results from the models. Further, the challenges faced are to be addressed for the successful implementation of the models in dentistry and to have long-term reliability on the models.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
The authors extend their appreciation to the Deanship of Postgraduate and Scientific Research at Dar Al Uloom University, Riyadh, KSA, for their support for this project.
1. Khanna SS, Dhaimade PA. Artificial intelligence: Transforming dentistry today Indian J Basic Appl Med Res. 2017;6:161–7
2. Wulff A, Montag S, Steiner B, Marschollek M, Beerbaum P, Karch A, et al CADDIE2-evaluation of a clinical decision-support system for early detection of systemic inflammatory response syndrome in paediatric intensive care: Study protocol for a diagnostic study BMJ Open. 2019;9:e028953
3. Deshmukh SV. Artificial intelligence in dentistry J Int Clin Dent Res Organ. 2018;10:47–8
4. Seol YJ, Kang HW, Lee SJ, Atala A, Yoo JJ. Bioprinting technology and its applications Eur J Cardiothorac Surg. 2014;46:342–8
5. Bas B, Ozgonenel O, Ozden B, Bekcioglu B, Bulut E, Kurt M. Use of artificial neural network in differentiation of subgroups of temporomandibular internal derangements: A preliminary study J Oral Maxillofac Surg. 2012;70:51–9
6. Bhan A, Goyal A, Chauhan N, Wang CW. Feature Line Profile Based Automatic Detection of Dental Caries in Bitewing Radiography 2016 International Conference on Micro-Electronics and Telecommunication Engineering (ICMETE). 2016;22:635–40
7. Russel S, Norvig P Artificial Intelligence: A Modern Approach. 20103rd ed. New Jersey Pearson Education
8. Kalappanavar A, Sneha S, Annigeri RG. Artificial intelligence: A dentist's perspective J Med Radiol Pathol Surg. 2018;5:2–4
9. Jackson J. Data mining: A conceptual overview Commun Assoc Inf Syst. 2002;8:267–96
10. Alabi RO, Elmusrati M, Sawazaki-Calone I, Kowalski LP, Haglund C, Coletta RD, et al Machine learning application for prediction of locoregional recurrences in early oral tongue cancer: A Web-based prognostic tool Virchows Arch. 2019;475:489–97
11. Lim K, Moles DR, Downer MC, Speight PM. Opportunistic screening for oral cancer and precancer in general dental practice: Results of a demonstration study Br Dent J. 2003;194:497–502
12. Yaji A, Prasad S, Pai A. Artificial intelligence in dento-maxillofacial radiology Acta Sci Dent Sci. 2019;3:116–21
13. Devito KL, de Souza Barbosa F, Felippe Filho WN. An artificial multilayer perceptron neural network for diagnosis of proximal dental caries Oral Surg Oral Med Oral Pathol Oral Radiol Endod. 2008;106:879–84