In the Eye of the Beholder: Changes in Perceived Emotion Expression after Smile Reanimation

Following is an abstract of this paper, which was published in Plastic and Reconstructive Surgery Journal in July 2019.

Introduction: Tools to quantify layperson assessments of facial palsy are lacking. Herein, artificial intelligence was applied to develop a proxy for layperson assessments, and compare sensitivity to existing outcome measures.

Methods: Artificially intelligent emotion detection software was used to develop the Emotionality Quotient (EmQ). EmQ was defined as the percent probability of perceived joy over the percent probability of perceived negative emotions during smiling, as predicted by the software. EmQ was used to analyze the emotionality of voluntary smiles or normal subjects and unilateral facial palsy patients before and after smile reanimation. EmQ was compared to oral commissure excursion and layperson assessments of facial palsy patients.

Results: In voluntary smiles of ten normal subjects, 100% joy and no negative emotion was detected (interquartile ranges (IQR) 0/1). Median preoperative EmQ of 30 facial palsy patients was +15/-60 (IQR 73/62). Postoperatively, median EmQ was +84/0 (IQR 28/5). In 134 smile reanimation patients, no correlation was found between postoperative oral commissure excursion and EmQ score. However, in 61 preoperative patients, a moderate correlation was found between layperson-assessed disfigurement and negative emotion perception (correlation coefficient 0.516, p<0.001).

Conclusion: Computer vision AI software detected less joy and more negative emotion in smiles of facial palsy patients compared with normals. Following smile reanimation, significantly more joy and less negative emotion were detected. EmQ was correlated with layperson assessments. The simplicity, sensitivity and objectivity of EmQ render it an attractive tool to serve as a potential proxy for layperson assessment, as ideal outcome measure in facial palsy.