(39) Artificial Intelligence Within Orthognathic Surgery Treatment Planning: Evaluation of a Pilot Machine-Based Learning Model

Abstract:

Introduction: Artificial intelligence (AI) is transforming various medical disciplines, including radiology and surgical planning. Despite AI’s broad applications, its integration into orthognathic surgery treatment planning remains limited. Traditional treatment planning methods rely on manual assessment and two-dimensional imaging, which can be time-consuming and prone to human error. Leveraging AI has the potential to streamline treatment planning, enhance diagnostic accuracy, and improve aesthetic and functional outcomes. This study aims to evaluate a machine-based learning tool designed for orthognathic surgery treatment planning, assessing its ability to generate precise and standardized treatment outputs.

Methods: An AI-driven orthognathic surgery treatment planning tool was developed to integrate intraoral scans and cone-beam computed tomography (CBCT) data. The model received inputs including standard cephalometric landmarks, midline discrepancy, maxillary central incisor show, maxillary cant, and the surgeon-set final occlusion. The tool then utilized these inputs to generate a standard orthognathic treatment plan predicting both hard and soft tissue changes. For purposes of case planning, the Barcelona line was utilized for esthetic position of the maxillary central incisors and the Gonzalez-Ulloa analysis was performed for chin position. A total of 10 previously planned orthognathic cases were evaluated by the AI software. The cohort was limited to Caucasian males that were planned for Le Fort osteotomies and bilateral sagittal split osteotomies. The AI-generated treatment plans were retrospectively compared to the movements dictated by the single experienced clinician at the initial time of treatment planning for variance.

Results: Preliminary analysis revealed that the AI-based model produced treatment plans that closely aligned with human-determined plans. However, notable variations were observed in occlusal plane rotation, suggesting that AI may offer unique insights into treatment modifications. Additionally, the AI tool demonstrated a high level of consistency in analyzing facial symmetry and occlusal relationships. While AI outputs generally matched traditional planning methods, discrepancies highlighted areas for further refinement and model training.

Discussion: The integration of AI into orthognathic surgery has the potential to improve treatment planning by increasing efficiency, reducing human error, and offering novel insights into jaw repositioning. The AI model utilized in this study showed promise in generating standardized and reliable treatment plans. However, it remains a supplemental tool rather than a replacement for expert clinical judgment. The authors are in no way recommending utilizing this tool for treatment planning alone. We do believe that this can both expedite treatment planning and allow for esthetic recommendations of movements.
One of the most significant applications of AI in orthognathic surgery is in preoperative planning. AI-powered simulations can provide enhanced visualizations of anticipated surgical outcomes, improving communication between surgeons, orthodontists, and patients. AI-powered integration can also allow for predictive recommendations for cases that may require atypical osteotomies or custom fixation/guides.

Future advancements should focus on expanding input parameters, including patient demographics such as race and sex, to refine predictions and improve personalization. Additionally, providers may have different opinions on the maxillary positioning or chin positioning and the model can easily be tailored to fit his or her preferences. These AI applications could extend beyond elective orthognathic procedures to include trauma and pathology cases, offering real-time assistance in complex surgical scenarios.

Conclusion: This study demonstrates that AI-based models have the potential to revolutionize orthognathic surgery treatment planning. While current models require further refinement, their ability to generate initial frameworks based on established esthetic and occlusal norms is promising. Long-term advancements should focus on incorporating additional clinical variables to optimize AI-driven treatment planning to streamline patient care and improve surgical outcomes.