Overview
The authors used machine learning to find objective measures linked to better cosmetic results. They compiled non-consecutive, selected photos of facelift, browlift and eyelid surgery from the American Society of Plastic Surgeons Before and After gallery. Two models scored perceived age and attractiveness before and after surgery, and odds ratio analyses tested which procedures and patient factors were linked to improvement.
Findings
- 676 patients, operated on by 65 surgeons in 22 states and provinces of the USA and Canada.
- Facelift 328 (48.5%), necklift 172 (25.4%), browlift 82 (12.1%), eyelid surgery 427 (63.2%).
- Older perceived age before surgery was linked to greater combined aesthetic benefit, regardless of procedure type: OR 1.04 (1.02-1.06), p < 0.001.
- Effects of procedures were additive. Facelift plus necklift plus browlift plus eyelid surgery showed the greatest combined benefit: OR 11.34 (1.84-103.34), p < 0.05.
- Lower attractiveness before surgery was also linked to greater benefit.
- Perceived age and attractiveness before surgery interacted in a complex, non-linear way.
- Level of evidence: III, retrospective multi-center cohort.
What it means for a patient
- By these AI measures, people who looked older before surgery gained the most.
- Combining procedures gave more measured benefit than single procedures.
- The confidence interval for the four-procedure combination is very wide (1.84-103.34), so the size of that benefit is uncertain.
- Limits: photos were non-consecutive, selected gallery entries, and the AI measures perceived age and attractiveness, not patient satisfaction.
Why this paper matters
It uses AI-scored perceived age and attractiveness as measures of facial surgery results across many surgeons. It links greater benefit to older perceived age and to combining procedures.
Terms
- Machine learning model: a computer program trained on many examples to score new images.
- Perceived age: how old a person looks, as judged here by the AI model.
- Composite aesthetic benefit score: a single score combining change in perceived age and attractiveness.
- Odds ratio (OR): a comparison of how likely an outcome is in one group versus another; above 1 means more likely.
- Multivariate analysis: a statistical test that accounts for several factors at once.
- Confidence interval: the range where the true value is likely to fall.
- Non-consecutive: patients not taken in order, so some cases may have been left out.