|Year : 2022 | Volume
| Issue : 1 | Page : 1-2
Artificial Intelligence in Oral and Maxillofacial Radiology: It is Here
Steven Richard Singer DDS
Department of Diagnostic Sciences, Division of Oral and Maxillofacial Radiology, Rutgers School of Dental Medicine, Newark, New Jersey, USA
|Date of Submission||26-May-2022|
|Date of Acceptance||27-May-2022|
|Date of Web Publication||05-Aug-2022|
Steven Richard Singer
Professor and Chair, Department of Diagnostic Sciences; Interim Director, Division of Oral and Maxillofacial Radiology, Rutgers School of Dental Medicine, 110 Bergen Street, PO Box 1709, Newark, New Jersey 07101-1709
Source of Support: None, Conflict of Interest: None
|How to cite this article:|
Singer SR. Artificial Intelligence in Oral and Maxillofacial Radiology: It is Here. J Orofac Sci 2022;14:1-2
Artificial intelligence (AI) has become firmly entrenched in dental research and is now finding its way into dental practice. AI in oral and maxillofacial radiology is presently an especially encouraging avenue of inquiry. Large data sets are being amassed by oral and maxillofacial radiologists to construct a massive and accurate diagnostic basis for convolutional neural networks to apply this accumulated experience to the interpretation of both intraoral and extraoral images.
How dentists will receive and use this new technology is still up in the air. Will it be perceived as an unwelcome usurper of our rightful place as expert clinicians, supplanting our years of experience parsing the contrasting radiodensities of periapical, bitewing, and panoramic images? Possibly a parlor game that we can use to impress patients with our skill to “beat” the computer? Perhaps we should be even more optimistic about its potential to be used as a serious diagnostic tool, as well as an in-house teacher of advanced radiographic interpretation and a backup for our tired eyes at the end of a long day in the office or clinic. It should be noted that “deep learning,” a subset of AI, can in certain fields, already delivers results that are at least on par with humans.
Development of radiographic interpretation systems faces many challenges. Perhaps most basic is the recognition of normal anatomic structures. A number of researchers have addressed this issue, using recognition of permanent teeth on panoramic radiographic images as the diagnostic challenge.- These studies have demonstrated that the performance of these systems is “comparable to the level of experts.” In addition to the obvious importance of correct labeling of teeth for diagnostic and communications purposes, this technology can be used to save valuable time that could be used for direct patient care.
A further benefit of AI can be the removal of human bias while interpreting images. Consider our possible disbelief at discovering that panoramic radiographs are indeed often comparable for interpreting common dental disease, or that extraoral bitewings may often be adequate for caries diagnosis. Perhaps our bias extends to the level of the individual patient, where we may be subconsciously less diligent where the patient has remained caries-free for many years. Convoluted neural networks have no such inherent bias and will analyze the images regardless of other potentially distracting factors.
While technical challenges still confront the developers of AI systems for oral and maxillofacial radiology, the deep convoluted networks are based upon the expert opinion of many highly regarded radiologists the world over. Therefore, when a single dentist views a radiograph, the interpretation, no matter how expert, is based on an n = 1, while AI interpretation contains, by definition, the expertise of a large number of Oral and Maxillofacial radiologists. It should be remembered, however, that the ultimate responsibility for correct interpretation still rests with the attending dentist. The acceptance of the results of such interpretation is still the patient’s choice. Perhaps a time will come when a patient may choose between a skilled human provider, a machine skill set, or possibly a combination of both. Would this potentially increase the cost of patient care? Improve the sensitivity and specificity of diagnosis? Only time will tell as the research related to radiology and AI continues to move forward.
| References|| |
Heo MS, Kim JE, Hwang JJ et al.
Artificial intelligence in oral and maxillofacial radiology: what is currently possible? Dentomaxillofac Radiol 2021;50:20200375. doi: https://doi.org/10.1259/dmfr.20200375
Tadinada A. Artificial intelligence, machine learning, and the human interface in medicine: is there a sweet spot for oral and maxillofacial radiology? Oral Surg Oral Med Oral Pathol Oral Radiol 2019;127:265–6. doi: https://doi.org/10.1016/j.oooo.2018.12.024
Estai M, Tennant M, Gebauer D et al.
Deep learning for automated detection and numbering of permanent teeth on panoramic images. Dentomaxillofac Radiol 2022;51:20210296. doi: https://doi.org/10.1259/dmfr.20210296
Tuzoff DV, Tuzova LN, Bornstein MM et al.
Tooth detection and numbering in panoramic radiographs using convolutional neural networks. Dentomaxillofac Radiol 2019;48:20180051. doi: https://doi.org/10.1259/dmfr.20180051
Prados-Privado M, García Villalón J, Blázquez Torres A, Martínez-Martínez CH, Ivorra C. A convolutional neural network for automatic tooth numbering in panoramic images. Biomed Res Int 2021;2021:3625386. doi: https://doi.org/10.1155/2021/3625386