Abstract
Introduction
Perceived age is an objective surrogate for facial rejuvenation, but comparative evidence
across facelift techniques using human and AI raters is limited.
Objectives and Hypotheses
This trial assessed whether facelift techniques differ in rejuvenation effect and
whether AI estimates align with human evaluations.
Study Design
Randomized clinical trial.
Methods
A total of 30 women (45–65 years) underwent rhytidectomy by deep plane, high SMAS,
or plication (n = 10 each). Standardized photographs were rated by 200 laypersons (9,000 evaluations)
and three AI models (180 evaluations). Primary outcome was change in perceived age
(Δ age); secondary analyses included technique comparison, AI accuracy, rater bias,
and human–AI correlation.
Results
All techniques significantly reduced perceived age, with no statistical difference
between groups. Amazon Rekognition and HowOldDoYouLook were more accurate. Human–AI
correlation was moderate (r = 0.41, p = 0.020). Raters under 30 underestimated age (p < 0.001).
Conclusion
Human and AI evaluations showed no technique differences, with AI estimates resembling
human assessments.
Keywords
rhytidoplasty - facial aging - artificial intelligence - esthetics - perception
