An Integrated AI Platform for Orthodontic Diagnosis and Treatment Planning

Uses for artificial intelligence (AI) have rapidly expanded within the orthodontic profession, offering new tools to improve diagnostic consistency, efficiency, and reproducibility.^1,2 Applications such as automated cephalometric analysis, skeletal maturation assessment, cone-beam computed tomography (CBCT) segmentation, and digital model interpretation allow complex datasets to be analyzed in a more standardized and objective manner, without replacing the orthodontist’s clinical judgment.^3-5

Beyond diagnosis, AI-driven systems are influencing treatment planning and decision-making. Machine-learning models can assist with extraction decisions, growth modification strategies, surgical case selection, and treatment simulation.^6,7 When combined with three-dimensional imaging and digital workflows, these systems enable visualization of proposed tooth movements, estimations of treatment duration, and more effective communication of anticipated outcomes to patients.

This article introduces DR ANNIE (Data Research Artificial Neural Network Intelligence Expert), an end-to-end AI-assisted platform that integrates orthodontic diagnosis, treatment planning, and outcome prediction into a unified clinical workflow. The conceptual framework and diagnostic components of the system are described, followed by a clinical case illustrating its application in contemporary orthodontic practice.

Conceptual Framework of the DR ANNIE Platform

Rather than functioning as a single algorithm, the DR ANNIE platform operates as a modular architecture in which multiple AI engines process raw clinical data, generate structured diagnostic outputs, and support treatment sequencing (Fig. 1).

The workflow begins with the acquisition of standardized digital orthodontic records, including clinical photographs, intraoral scans, and panoramic and lateral cephalometric radiographs, as well as anteroposterior cephalometric and CBCT imaging when indicated. These inputs are processed through a series of specialized AI modules, each responsible for a specific diagnostic task such as landmark detection, skeletal classification, segmentation of dental and skeletal structures, and quantitative assessment of occlusal and anatomical relationships. Additional modules can be used to analyze clinical photographs for malocclusion, treatment needs, caries, periodontal disease, and restorations⁸; segment panoramic radiographs⁹; assess the TMJ¹⁰; and estimate case difficulty and treatment length.

Key findings influencing treatment decisions included the growth pattern, an ovoid lower archform, a lower arch-length discrepancy of 3.7mm, an absence of leeway space, and a lack of indication for orthopedic or auxiliary appliances. The AI-supported decision analysis recommended a nonextraction approach in both arches, with space management achieved through minor posterior distalization, limited incisor repositioning, and transverse coordination with buccal segment alignment (Fig. 7A). Alternative approaches considered for this patient included limited extraction therapy and conventional fixed appliance therapy. The final decision was made by the treating orthodontist after considering the patient’s esthetics, growth pattern, and eruption status, as well as biomechanical feasibility. Clear aligner therapy was selected based on the doctor’s preference.

The platform then generated a step-by-step treatment staging sequence for each arch, with an estimated overall treatment duration of 16-17 months (Fig. 7B). In both arches, the early phases emphasized posterior tooth distalization and incisor vertical control, followed by buccal segment alignment and refinement of archform.

The platform’s LLM-based narrative translated technical staging into patient-friendly language, describing month-by-month objectives for each arch (Fig. 8). This narrative was reviewed and edited by the orthodontist before it was used in discussions with the patient and parents and for communication with the aligner laboratory.

The clinician-approved staging was exported to the aligner simulation (Fig. 9).

After 11 months of clear aligner therapy, the observed tooth positions corresponded closely with the predicted stages (Fig. 10).

Discussion

AI platforms such as DR ANNIE are designed to integrate into existing orthodontic workflows. In routine practice, they can be introduced at the diagnostic and consultation stages using records that meet the established standard of care. By automating time-intensive analytical tasks, AI enables orthodontists to focus more on clinical decision-­making, patient communication, and treatment execution. AI systems may also reduce variability between providers within group practices or multilocation clinics.

Visual representations of malocclusion severity, treatment objectives, and step-by-step staging, reinforced by clinician-reviewed narratives, allow patients and parents to better understand the rationale behind recommended treatment. This improved understanding has the potential to increase case acceptance, help manage expectations, and improve compliance.

We recommend gradual adoption of an AI platform, beginning with diagnostic assistance and case visualization before progressing to treatment staging and outcome prediction. Such an approach will enable the orthodontist and staff to build trust in the system while preserving effective workflows.

AI-based systems do have certain limitations. First, their detections and recommendations can be inaccurate and must be verified and interpreted by the orthodontist. In addition, although several individual components of DR ANNIE have been validated, the fully integrated system has yet to undergo an end-to-end validation.

As AI models continue to evolve, further applications are expected to include more refined outcome prediction, integration of longitudinal treatment data, and personalized risk assessment. The most effective future applications of AI in orthodontics will be those that can augment clinical expertise, enhance patient-centered care, and support evidence-based decision-making, without replacing the clinician’s judgment.