a clinical exam, which is fine for epidemiological screening. The issue is that if you don’t build your AI database on diagnoses from clinical examinations, you are more prone to false positives. What may appear carious is, in reality, not an active lesion. This could lead to overtreatment, which was one of the concerns that came out of our research. SS: I’d like to reinforce what Dr. Moharrami just mentioned. For these tools to be effective in a dental office, they must be based on a diverse range of source materials, with dental caries in the training photos diagnosed through clinical examinations. While the AI tools used in these studies may be beneficial in a public health setting such as school screening, they are not yet suitable for use in a dental office. As a quaternary prevention principle, one needs to be cautious that any tool, such as AI, should not lead to over treatment. MM: There is still a lot to explore in this field of study, and the literature documented other applications of AI in oral photographs, such as screening of precancerous lesions and certain orthodontic outcomes. For the next projects, I’m interested in looking at whether AI and oral photographs could be effective at detecting dental plaque and gingivitis. We are also exploring the idea of original research using data from our clinic at the U of T faculty of dentistry. Dentistry hasn’t been able to leverage AI as quickly as medicine, partly because we don’t have the same levels of readily available data in oral health. The biggest studies that we included in our review had only about 1,000 oral photographs, which is relatively small compared to other fields. To build effective AI tools, we need to collect more high-quality data and make it more accessible. If I could advocate for something as a researcher, it would be that we simply need more data in oral health. SS: With the launch of the federal government’s Canadian Dental Care Plan, there is great potential for progress. Dentists across Canada, providing care to millions of people eligible under the plan, are doing clinical examinations, taking radiographs and oral photographs, and making diagnoses. The funding agency, the plan administrator, and dental associations can join hands to develop a databank of anonymized data, which could serve as a significant resource for further enhancing this kind of AI tool. MM: At this point, a lot of the hard work has already been done. AI infrastructure and algorithms exist, and nowadays, we have the computational power to train and deploy AI models. Now, it’s all about having the right data so that we can build tools that will be useful in improving people’s oral health. Reference 1. Moharrami M, Farmer J, Singhal S, Watson E, Glogauer M, Johnson AEW, Schwendicke F, Quiñonez C. Detecting dental caries on oral photographs using artificial intelligence: A systematic review. Oral Dis [Internet]. 2023;30(4):1765-83. There is still a lot to explore in this field of study, and the literature documented other applications of AI in oral photographs, such as screening of precancerous lesions and certain orthodontic outcomes. 22 | 2025 | Issue 1 Issues and People
RkJQdWJsaXNoZXIy OTE5MTI=