Automated Spontaneity Assessment after Smile Reanimation: An Artificial Intelligence Approach
Joseph R Dusseldorp, MBBS, MS, FRACS1, Diego L Guarin, PhD1, Martinus M van Veen, MD1, Nate Jowett, MD2 and Tessa A Hadlock, MD3, (1)Harvard University, Boston, MA, (2)Facial Nerve Center - Dept. of Otolaryngology, Harvard Medical School / Massachusetts Eye and Ear, Boston, MA, (3)Facial Nerve Center - Dept. of Otolaryngology, Harvard Medical School / Massachusetts Eye and Ear Infirmary, Boston, MA
INTRODUCTION: Re-creation of a spontaneous, emotional smile remains an overriding goal of smile reanimation surgery. However, the ideal innervation strategy remains unknown. An automated machine-learning tool was developed to compare spontaneous smiling in cross facial nerve graft, masseteric nerve, and dually-innervated free gracilis transfers.
MATERIALS AND METHODS: Validated humorous videos were used to elicit spontaneous smiles. Automated facial landmark recognition (Emotrics) and emotion detection software were used to analyze video clips of spontaneous smiling in nine normal subjects, and 43 facial palsy patients. Emotionality quotient (EmQ: probability of perceived joy / probability of perceived negative emotion) was used to quantify the ability of spontaneous smiles to express joy.
RESULTS: Spontaneous smiles of normal subjects exhibited median 100% joy and almost no negative emotion (EmQ score +100/0). Spontaneous smiles of facial palsy patients after smile reanimation, using cross facial nerve graft, masseteric nerve, and dual innervation, yielded EmQ scores of median +82/0, 0/-48,and +10/-24 respectively (joy p = 0.006, negative emotion = p = 0.034). The main difference was found to be between the cross face and masseteric nerve driven gracilis FFMT (joy p = 0.001, negative emotion p = 0.008).
CONCLUSION: Computer vision software can objectively quantify differences between various reinnervation strategies in facial reanimation. Cross-facial nerve graft- and dually-innervated gracilis achieved greater degrees of improvement in emotionality during spontaneous smiling, whilst masseteric alone was significantly worse. This automated system for quantification of spontaneous smiling from standard video clips would facilitate blinded, multi-center comparisons of spontaneity outcomes.
Back to 2020 Abstracts