Assessment of Facial Synkinesis Treatment with Botulinum Toxin Using Automated Analysis of Facial Expression: A Pilot Study

Background: The lack of quantification of facial function makes outcome analysis in facial synkinesis challenging. Learning/Study Objective: To compare facial expressions in patients with synkinesis before and after botulinum toxin treatment as measured by the standard electronic clinician-graded facial function scale (eFACE) compared with an automated emotional expression software (FaceReader). Design Type: Retrospective cohort. Methods: Patients underwent botulinum toxin treatment for synkinesis with injections in the platysma, depressor labii inferioris (DLI), depressor anguli oris (DAO), mentalis, orbicularis, glabella, forehead, and zygomatic region. Standardized photographs were taken 30.9 ± 5.55 days after injection and graded with eFACE and FaceReader. Wilcoxon matched-pairs signed-rank test compared the pre- and post-therapeutic scores. Spearman's correlation compared eFACE and FaceReader scores. Results: Three men and 15 women (mean age 48.33 ± 3.17 years) with unilateral synkinetic facial paralysis (9 left, 9 right, most commonly oculo-oral and/or oro-ocular [94%], mean duration 9.72 ± 2.87 years, most commonly idiopathic [50%]) were included. The eFACE synkinesis scores significantly improved (p < 0.05. The FaceReader "happy" score increased at rest, the "disgusted," and "scared" score while smiling with exposed teeth decreased (p < 0.05). We observed positive correlations between the platysmal, ocular, and total eFACE synkinesis scores and FaceReader "happy" score (r = 0.59, p = 0.019; r = 0.22, p = 0.404; r = 0.33, p = 0.211, respectively). Conclusions: AI-driven analysis of facial synkinesis and botulinum toxin treatment shows some correlation with the eFACE scores but is not yet reliable.