Perceived Age and Gender Perception Using Facial Recognition Software Following Facial Feminization Surgery

Abstract

Measures of success for facial feminization surgery (FFS) have previously included improved rates of external gender perception as female and patient-reported outcome measures. In this study, we used artificial intelligence facial recognition software to objectively evaluate the effects of FFS on both perceived gender and age among male-to-female transgender patients, as well as their relationship with patient facial satisfaction. Standardized frontal preoperative and postoperative images of 27 transgender women undergoing FFS were analyzed by Amazon’s AI facial recognition software to determine gender, femininity confidence score, and perceived age. Female gender-typing, improvement in gender-typing (preoperatively to postoperatively), and femininity confidence scores were analyzed. To assess patient satisfaction, FACE-Q modules were completed postoperatively. Preoperatively, FFS images were perceived as female 48.1% of the time, and postoperatively, this improved to 74.1% (P=0.05). Femininity confidence scores improved from a mean score of 0.04 preoperatively to 0.39 postoperatively (P=0.003). FFS was associated with a decrease in perceived age relative to the patient’s true age (−2.4 y, P<0.001), with older patients experiencing greater reductions. Pearson correlation matrix found no significant relationship between improved female gender typing and patient facial satisfaction. Undergoing surgery at a younger age was associated with higher overall facial satisfaction (r=−0.6, P=0.01). Transfeminine patients experienced improvements in satisfaction with facial appearance, perceived gender, and decreases in perceived age following FFS. Notably, patient satisfaction was not directly associated with improved AI-gender typing, suggesting that other factors may influence patient satisfaction.

Applications of artificial intelligence in facial plastic and reconstructive surgery: a systematic review

“…Postoperatively, the accuracy of sex-typing increased 26% from 48.1 to 74.1%, with younger patients (20–30 years old) identified as female 100% of the time. Additionally, femininity confidence scores improved from 0.04 to 0.39 postoperatively and average perceived age decreased by 2.4 years [14].…”

“…Patients who appeared older or less attractive at baseline had the largest decrease in perceived age and increase in perceived attractiveness, respectively [15]. Alper et al [14] used Rekognition, an Amazon Web Services platform, to assess changes in sextyping, female confidence score, and apparent age before and after facial feminization surgery (FFS). Postoperatively, the accuracy of sex-typing increased 26% from 48.1 to 74.1%, with younger patients (20-30 years old) identified as female 100% of the time.…”

“…Artificial intelligence has shown promise in preoperative surgical planning, cross-sectional imaging analysis, facial palsy assessment, and developing inclusive sex, race, and aging evaluation tools [4 when benchmarking outcomes of FPRS procedures [10][11][12][13][14][15]. In a nonclinical setting, artificial intelligence has been used as a tool to drive research, supplement medical education, and provide digestible insights into care for patients [16][17][18][19][20][21].…”

“…Artificial intelligence has also demonstrated utility as an intraoperative tool by providing guidance through augmented reality or mapped neural network tools [8 ▪▪ ,9]. In a postoperative setting, artificial intelligence can provide objectivity when benchmarking outcomes of FPRS procedures [10–15]. In a nonclinical setting, artificial intelligence has been used as a tool to drive research, supplement medical education, and provide digestible insights into care for patients [16–21].…”

Artificial Intelligence as a Triage Tool during the Perioperative Period: Pilot Study of Accuracy and Accessibility for Clinical Application

“…The current literature using AI describes various attempts to integrate various functions into preoperative surgical planning, patient education, and assessment of outcomes. [9][10][11][12][13][14][15][16] No study to our knowledge has assessed the feasibility of a patient interactive AI program for answering questions, troubleshooting complications, and providing clinical recommendations during the perioperative period.…”