Machines can read X-rays within seconds. They can detect anatomical landmarks on cephalometric images without fatigue or mood-related bias. But can they replace the trained eye of an orthodontic specialist? A scoping review involving drg. Rellyca Sola Gracea, Sp.RKG. Subsp.RDP(K), and colleagues from overseas provides a cautious answer: not yet.
The study, entitled “Artificial Intelligence for Orthodontic Diagnosis and Treatment Planning: A Scoping Review,” was published in the Journal of Dentistry in 2024. The research team, consisting of Gracea, Winderickx, Vanheer, and colleagues, conducted a systematic search of three major scientific databases: PubMed, Web of Science, and Embase. Of the hundreds of articles screened, 71 studies were ultimately included in the review.
Three Battlegrounds for AI in Orthodontic Practice
The study mapped the applications of artificial intelligence in orthodontics into three major domains: diagnosis (29 studies), anatomical landmark identification (20 studies), and treatment planning (22 studies). These numbers are more than mere statistics. They reflect where researchers around the world are currently directing their attention.
The most extensively studied domain was anatomical landmark detection—the ability of AI to identify specific reference points on cephalometric images. In conventional clinical practice, this task requires a high level of concentration and extensive experience. AI offers speed and consistency, two areas in which humans can be disadvantaged when dealing with large volumes of cases.
Yet the most crucial finding was not about what AI can do, but what it still cannot do.
“AI shows potential for improving time efficiency and reducing operator variability, but its accuracy and reliability have not consistently surpassed those of expert clinicians.” — Gracea RS, Winderickx N, Vanheer M, et al., Journal of Dentistry, 2024
Efficiency, Yes; Supremacy, Not Yet
Imagine two scenarios in a clinical setting: an experienced orthodontist sits in front of a monitor, carefully examining a cephalogram for twenty minutes. Beside the orthodontist, an AI system completes the same analysis in two minutes. Which one is more accurate?
According to this review, the answer still depends on the context. AI excels in speed and can reduce inter-operator variability, a genuine issue in clinical practice in which two clinicians may produce different interpretations of the same image. However, when it comes to absolute accuracy and the ability to integrate multiple clinical variables holistically, humans still have the advantage.
This is what makes the review particularly balanced. It does not succumb to excessive optimism about AI. Orthodontic treatment planning is not merely a matter of facial and dental geometry. Psychological factors, aesthetic preferences, periodontal conditions, and projected skeletal growth must all be considered together. To date, no AI system has been able to integrate all of these factors automatically and reliably.
Human Oversight: A Non-Negotiable Requirement
The researchers emphasized one point repeatedly throughout the review: human oversight remains essential at every stage of AI use. This is not a mere formality but a safeguard for patient safety.
An AI system incorrectly identifying a single landmark may seem insignificant on a screen. However, a small error can lead to an inaccurate treatment plan, inappropriate tooth movement, and even an unfavorable clinical outcome. No matter how sophisticated an AI system may be, it still lacks the ability to recognize its own limitations.
The study therefore concluded with a call for further development. This includes algorithm optimization, validation across more diverse populations, and, most ambitiously, progress toward truly patient-specific and automated treatment planning. It is an aspiration that has not yet been achieved, but it is not necessarily impossible.
The dental profession is learning to work alongside AI rather than compete against it. And perhaps that is the most sensible way forward.
Source DOI: https://doi.org/10.1038/s41415-024-8237-3
Authors: Anny Anggraini; drg. Achmad Zam Zam Aghasy, M.Kes.
Photo: Freepik