Exploring Facial Asymmetry Metrics Correlated with Preoperative SCHNOS Scores Using Artificial Intelligence.

Abstract

Background: Facial asymmetry is often overlooked in evaluations of nasal function and aesthetics, despite its potential impact on assessments in facial plastic surgery. In this study, artificial intelligence tools were used to identify facial asymmetry metrics that correlate with both nasal function and aesthetic measures evaluated by preoperative Standardized Cosmesis and Health Nasal Outcomes Survey (SCHNOS) scores.

Methods: Two facial landmark detection models were applied to frontal plain facial images of 1523 patients to extract 506 fiducial points. From these, more than 64 million facial elements were computed, including point-to-point and point-to-line distances. Asymmetry indices were calculated based on each element with its mirrored counterpart. Spearman correlation coefficients were used to assess associations between these asymmetry metrics and 13 outcome scores.

Results: Facial elements correlated with SCHNOS-nasal obstruction demonstrated modest but statistically significant Spearman correlations (0.185 to 0.224; P < 10 -11 ), particularly those capturing vertical facial height differences relative to a horizontal reference line between the nasal tip and ear base. No meaningful correlations were observed with SCHNOS-nasal cosmesis scores.

Conclusions: These findings suggest that vertical midfacial asymmetry may affect nasal function, whereas facial asymmetry has minimal influence on patients' perception of nasal aesthetics. The study also underscores the potential of artificial intelligence-based facial analysis as a valuable tool in rhinoplasty evaluation.