"Objectifying Outcomes in Facial Aesthetic Surgery Using Artificial Intelligence"

Choosing which procedure or combination of procedures is appropriate remains a daunting task for patients and surgeons alike. The goal of this study was to leverage machine learning models (MLMs) to find objective AI-derived parameters associated with improved cosmetic outcomes in patients undergoing aesthetic procedures.

A dataset from the American Society of Plastic Surgeons (ASPS) Before and After photos was compiled for facelift, browlift, and eyelid surgeries using non-consecutive, curated photographs of all available entries. Two MLMs were trained to detect perceived age and attractiveness using "in the wild" reference images. Pre- and post-operative photographs were used to generate scores before and after surgery per patient. Uni- and multi-variate odds ratio (OR) analyses were performed to interrogate procedures and demographic factors associated with improvements in age perception, attractiveness, and a composite aesthetic benefit score.

676 patients operated on by 65 surgeons from 22 different states/provinces in the USA and Canada were included in the analyses. 328 (48.5%), 172 (25.4%), 82 (12.1%), and 427 (63.2%) underwent facelift, necklift, browlift, or eyelid surgeries, respectively. On multivariate analysis [OR (CI), p-value] older perceived preoperative age was associated with a greater combined aesthetic benefit independent of procedure type [1.04 (1.02-1.06), p < 0.001]. The effects of surgery were found to be additive, with patients undergoing combined procedures (facelift+necklift+browlift+eyelid) demonstrating the greatest composite aesthetic benefit [11.34 (1.84-103.34), p < 0.05].

Machine learning can be used to more objectively assess aesthetic facial surgery using perceived age and attractiveness as novel preoperative markers. Perceived older preoperative age, decreased preoperative attractiveness, and increased number of concurrent procedures were associated improved cosmetic benefit. However, the interaction between perceived preoperative age and attractiveness demonstrated a complex, non-linear relationship when optimizing for postoperative aesthetic benefit.Clinical Question/Level of Evidence: Level III- Retrospective Multi-Center Cohort Study.