Study selection and search results
Electronic searches conducted via PubMed and Scopus identified a total of 776 studies, of which 51 duplicate records were deleted. Of the remaining 725 studies, 665 were excluded based on title irrelevant. This was followed by 60 studies undergoing abstract and full-text screening, and 34 studies from electronic search. Additional manual searches using Google Scholar yielded 123 studies, of which 10 were included. In total, this review included 44 studies.
Research summary characteristics
The studies included in this review were published between 2011 and 2024, with most studies reporting sample sizes ranging from 10 to 68 patients. Of the 44 studies analyzed, 26 studies [12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37] It was retrospective, while seven studies. [38,39,40,41,42,43,44] I used a comparison design. The remaining studies consisted of three observations. [45,46,47],5 verification [48,49,50,51,52]2 sections [53, 54]and one prospective study [55] (Supplementary Table). When it comes to using AI-based applications, diagnostics are the most frequently employed tool and was used in 26 studies. [13,14,15,16, 18,19,20,21,22,23,24, 27, 28, 30, 32, 33, 35,36,37, 41, 43, 51, 52, 55]cephx continues in 16 studies [12, 17, 25, 26, 29, 31, 34, 38, 40, 44, 45, 47,48,49,50, 54]. Furthermore, two studies [39, 53] Using Denti.ai, smile.ai [42] And a smile designer [46] Each was reported in one study. The level of development of research using the Fryback and Thornbury models was varied. Particularly, 27 studies [13,14,15,16, 19, 21,22,23, 25, 26, 28, 30, 31, 33, 36, 39,40,41, 43,44,45,46, 50,51,52,53, 55] We achieved a level 2 deployment that evaluates AI software and tests its effectiveness without any actual deployment. However, six studies [20, 24, 38, 42, 47, 49] Achieve Level 1 and 6 other studies [17, 18, 34, 35, 37, 48] I reached level 3 and 3 studies [12, 29, 54] I've reached level 4. Only one study achieved level 5 [32] Level 6 each [27]respectively, assess the effectiveness of AI-based applications in clinical workflows. Regarding discipline, 20 studies had AI-based applications focused on the field of orthodontics [12, 17, 21, 25, 26, 29,30,31, 34, 36, 38, 40, 42, 44, 45, 47,48,49,50, 54]followed by seven studies of endodontic therapy. [13, 33, 35, 39, 41, 43, 52]Six Dental Radiology Research [14, 15, 22, 27, 32, 53]and five studies of oral and maxillofacial surgery. [16, 18, 23, 24, 28]. However, there were only two prosthetic research studies. [19, 46] and periodontal disease [20, 51]and only one restorative dentistry research [37] and oral medicine [55].
I'll contact the company
To assess effectiveness and deployment, we sought to gather detailed information about AI-based dental applications by reaching out to various companies to provide AI solutions to dental practitioners as shown in Table 1, as shown in Table 1. Pearl© and Dentalxr.ai©. Despite multiple inquiries via email, only the developers of Cephx Company responded that they would instead offer a free trial to complete the access needed to evaluate the actual deployment of the application.
Deployment levels refer to various stages of research that specifically illustrate applications in real-world settings, ranging from technical reports (level 1) to assessments of broader social impacts from levels 5-6. The multi-level framework shown in Figure 1 lives primarily in research focusing on the effectiveness of deployment level 1: technical effectiveness and level 2: diagnostic accuracy to assess AI in dental treatment. These studies primarily include conceptual principles and validation that evaluates the performance and diagnostic accuracy of AI algorithms. Level 1 studies highlight parameters such as algorithm accuracy, sensitivity to imaging artifacts, and computational efficiency, while Level 2 studies focus on metrics such as sensitivity, specificity, and positive predictors. In contrast, relatively few studies have explored higher levels of efficacy, such as Level 3: Effectiveness of Diagnostic Thinking, Level 4: Effectiveness of Treatment, Level 5: Effectiveness of Patients, and Level 6: Social Efficacy.
