Can Artificial Intelligence Replicate Clinical Judgment in Aesthetic Surgery? A Comparison of ChatGPT and Plastic Surgeons' Recommendations.
Since the emergence of the Chat Generative Pre-Trained Transformer (ChatGPT), artificial intelligence (AI) has change how medical information is accessed, processed, and utilized. However, its use in clinical environments, particularly in plastic surgery, poses ethical, medical, and safety challenges, especially when it is employed to generate aesthetic recommendations without appropriate professional supervision.
A descriptive analysis was conducted using standardized medical photographs of 20 female patients under 50 years of age from the author's clinical database. Frontal projection images (thorax and abdomen) were uploaded to the ChatGPT-4 platform with the prompt: "What physical changes do you suggest?" and subsequently evaluated by board-certified plastic surgeons. The objective was to compare AI-generated recommendations with those provided by human specialists and to analyze the risks associated with using algorithms to validate or reinforce aesthetic standards.
In 18 of the 20 patients (90%), the AI suggested at least one surgical procedure, most commonly mastopexy (55%), classic abdominoplasty (55%), and liposculpture (20%). Combined breast and abdominal procedures were recommended in 15 cases (75%), whereas only 2 patients (10%) were advised to maintain their current physical condition. Overall similarity between AI and human recommendations was low (mean Jaccard index 0.1-0.2), with wide interpatient variability. Additionally, AI tended to recommend an equal or greater number of procedures compared with surgeons (median - 1, p < 0.05).
Although artificial intelligence represents a promising tool in medicine, AI tends to over-recommend aesthetic procedures without considering the reason for consultation, timing, or patient comorbidities. Its indiscriminate use in the aesthetic domain may contribute to reinforcing distorted perceptions of body image. Lacking clinical judgment, empathy, and sociocultural context, the application of AI in plastic surgery must be carefully regulated and guided by ethical principles.
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 .
A descriptive analysis was conducted using standardized medical photographs of 20 female patients under 50 years of age from the author's clinical database. Frontal projection images (thorax and abdomen) were uploaded to the ChatGPT-4 platform with the prompt: "What physical changes do you suggest?" and subsequently evaluated by board-certified plastic surgeons. The objective was to compare AI-generated recommendations with those provided by human specialists and to analyze the risks associated with using algorithms to validate or reinforce aesthetic standards.
In 18 of the 20 patients (90%), the AI suggested at least one surgical procedure, most commonly mastopexy (55%), classic abdominoplasty (55%), and liposculpture (20%). Combined breast and abdominal procedures were recommended in 15 cases (75%), whereas only 2 patients (10%) were advised to maintain their current physical condition. Overall similarity between AI and human recommendations was low (mean Jaccard index 0.1-0.2), with wide interpatient variability. Additionally, AI tended to recommend an equal or greater number of procedures compared with surgeons (median - 1, p < 0.05).
Although artificial intelligence represents a promising tool in medicine, AI tends to over-recommend aesthetic procedures without considering the reason for consultation, timing, or patient comorbidities. Its indiscriminate use in the aesthetic domain may contribute to reinforcing distorted perceptions of body image. Lacking clinical judgment, empathy, and sociocultural context, the application of AI in plastic surgery must be carefully regulated and guided by ethical principles.
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 .
Authors
González González, Danilla Danilla, Las Heras Las Heras, BenĂtez BenĂtez, Albornoz Albornoz, Konhenkampf Konhenkampf, DomĂnguez DomĂnguez
View on Pubmed