NESPS Home  |  Past Meetings
The Northeastern Society of Plastic Surgeons

Back to 2020 Abstracts


Using Artificial Intelligence to Analyze Facial Action Units Following Facial Reanimation Surgery
Thanapoom Boonipat, MD1, Jason Lin, BS1, Malke Asaad, MD1, Mitchell Stotland, MD2, Samir Mardini, MD1.
1Mayo Clinic, Rochester, MN, USA, 2Sidra Medical, Doha, Qatar.

BACKGROUND: Facial expressions can be analyzed systematically using the Facial Action Coding system which links discrete facial action units to specific emotions 1,2. This study evaluates the use of a machine learning technology to directly measure facial action unit and emotional expression both before and after facial reanimation surgery. METHODS: Fifteen study subjects with facial palsy were evaluated both before and after undergoing cross facial nerve graft and free gracilis muscle transfer. Video footage was obtained of subjects with their face both in repose and with a posed, closed-lip smile. The video data was then analyzed using the Noldus FaceReader™ software application to measure the 28 action units and the happy and sadness emotion detected within each clip. RESULTS: During smile, activation of the action units corresponding to happy emotions increased from average intensity of 1.5/4 pre to 2.71/4 post operatively and 1.0/4 pre to 1.83/4 postoperatively in the lip corner puller and cheek raiser AUs, respectively. This corresponds to an increase in overall happy emotion detected from 13% to 42% (p<0.0001).Conversely, the “lip corner depressor” action unit, associated with sad emotions, decreased from average of 2/4 pre to 0/4 postoperative during smile. The sad emotion detected decreased from 15% to 9% (p=0.092). The other two action units associated with sad emotions (“inner brow raiser “and “brow lowerer”) did not show any change pre- and post-operatively value of 1/4. CONCLUSIONS: This study provides the first proof of concept for the use of a machine learning software application to objectively detect facial units changes and quantify facial expression before and after surgical reanimation.


Back to 2020 Abstracts