AutoFFS: Adversarial Deformations for Facial Feminization Surgery Planning
Researchers' novel framework transforms skull shapes by attacking sex classifiers, creating data-driven surgical blueprints.
A research team from the University of Basel and University Hospital Basel has published a groundbreaking paper on arXiv introducing AutoFFS, a data-driven framework for planning facial feminization surgery (FFS). The system addresses a critical gap in transgender healthcare: current FFS planning relies heavily on subjective clinical assessment, lacking quantitative and reproducible anatomical guidance. AutoFFS innovates by generating counterfactual skull morphologies through adversarial free-form deformations, effectively creating a data-driven blueprint for surgical modification.
The technical core of AutoFFS involves performing a deformation-based targeted adversarial attack on an ensemble of pre-trained binary sex classifiers that have learned sexual dimorphism from data. By 'fooling' these classifiers, the system transforms an individual's 3D skull scan toward the target sex morphology, producing a specific, quantitative surgical target. The approach was validated through both classifier-based evaluation and a human perceptual study, confirming the generated morphologies exhibit the intended female characteristics. This represents a major shift from art to science in surgical planning for a historically overlooked patient group, potentially improving outcomes, consistency, and accessibility of gender-affirming care.
- AutoFFS performs adversarial attacks on sex classifiers to transform 3D skull scans toward female morphology.
- The framework provides quantitative, data-driven surgical plans, moving beyond subjective clinical assessment.
- Validated by classifier metrics and human perceptual studies, confirming target sex characteristics in outputs.
Why It Matters
Introduces objective, reproducible planning for gender-affirming facial surgery, potentially standardizing care and improving patient outcomes.