An Artificial Intelligence Tool Used for Patient Selection in Cosmetic Surgery

Background: Cosmetic surgery is becoming increasingly popular, but not all patients are suitable candidates. A thorough preoperative assessment and careful patient selection are essential for achieving successful outcomes. Given the challenges plastic surgeons face in conducting psychological assessments, we believe that artificial intelligence (AI) has the potential to assist in this critical task of patient selection. Methods: Based on previous research findings and the expertise of six senior plastic surgeons, we developed an AI-based patient selection tool using the XGBoost algorithm. We then enrolled 37 patients, who were independently evaluated by both the AI system and two senior plastic surgeons. This clinical validation process was designed to assess the accuracy and reliability of the AI tool in identifying suitable candidates for cosmetic surgery. Results: Clinical validation with 37 patients demonstrated a high level of agreement between the AI system and the surgeons in patient assessment and selection. Both the AI and the surgeons agreed that 73% of patients were suitable for surgery, while 27% were not. Quantitatively, there was a strong positive correlation between the AI and the surgeons’ assessments, with correlations of r = 0.907 (95% CI = [0.837, 0.949], p < 0.001) for the average surgeon rating, r = 0.909 (95% CI = [0.824, 0.955], p < 0.001) for Surgeon A, and r = 0.879 (95% CI = [0.798, 0.930], p < 0.001) for Surgeon B. Conclusion: We’ve developed the first smartphone-based AI tool designed to evaluate patients prior to cosmetic surgery. This practical tool enables surgeons to make wise and effective patient selection. The AI tool demonstrated strong performance in our clinical validation. Level of Evidence IV: This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266.