Readspeaker Menü

Facial Paresis Analysis

Oliver Mothes, Martin Thümmel and Joachim Denzler


Motivation

The human face is a complex and diverse part of the human body. All facial motions for individual mimics and emotional expressions are possible using 43 different muscles controlled respectively by one facial nerve on both halves of the face. Unfortunately, this nerve can be injured by external or internal factors, which leads to a facial palsy. A unilateral facial palsy essentially leads to a one-sided motor dysfunction of the facial muscles, a movement disorder. The resulting defective eye closure leads to the drying out of the eye, eye inflammation and visual impairment.  The paralysis of the perioral musculature hinders regular dient and speech. Medical experts evaluate facial palsy in a subjective manner. But ideally, these functional deficits should be measured objectively by exploiting machine learning methods. The following projects tackle this task.

Project: Bridging the Gap - Mimics and Muscles

The aim of the project is the synthesis of the facial surface model with a model of the facial muscles. High-resolution 3D video sequences, which are recorded synchronously together with electromyography sequences, are intended to facilitate understanding of the facial muscles. The resulting model should be used for an automatic objective evaluation of facial palsy.


Project: Irritationsfreies und emotionssensitives Trainingssystem (IRESTRA)

The interaction between humans and machines is an ongoing research field in many scientific disciplines. To develop an optimal strategy for machines to interact with humans the interactions between humans can be used to detect signals for the mood of the interaction partner and react in an adequate way. Especially elderly people can benefit from an easy and intuitive handling of technical devices. The aim of the project is to develop an irritation free and non invasive emotion sensitive training device for elderly people and Facial Paresis patients. Older persons can use this system for memory training (e.g. person recognition), whereas paresis patients can do their daily facial training sessions with a direct feedback from the machine and motivation if necessary. Within this project these tasks are tackled by combining medical, psychological and neurological knowledge with machine learning and computer vision techniques both in 2D and 3D in coorperation with the Fraunhofer Institute for Applied Optics and Precision Engineering IOF, the Jena University Hospital (namely the Institute of Psychosocial Medicine and Psychological Therapy, the Department of Neurology and the Department of Otolaryngology) and the Department for General Psychology from the Friedrich Schiller University Jena.


Facial Paresis Index Prediction

  Framework for Facial Paresis Index Prediction

[Mothes19:ORL]

Oliver Mothes and Luise Modersohn and Gerd Fabian Volk and Carsten Klinger and Otto W. Witte and Joachim Denzler and Olando Guntinas-Lichius. Automated objective and marker-free facial grading using photographs of patients with facial paksy. European Archives of Oto-Rhino-Laryngology.2019.


[Modersohn16:FPAAM]

Luise Modersohn and Joachim Denzler. Facial Paresis Index Prediction by Exploiting Active Appearance Models for Compact Discriminative Features. International Conference on Computervision Theory and Application (VISAPP) 2016. 271-278. [pdf] [bib]
Summary In the field of otorhinolaryngology, the dysfunction of the facial nerve is a common disease which results in a paresis of usually one half of the patients face. The grade of paralysis is measured by physicians with rating scales, e.g. the Stennert Index or the House-Brackmann scale. In this work, we propose a method to analyse and predict the severity of facial paresis on the basis of single images. We combine feature extraction methods based on a generative approach (Active Appearance Models) with a fast non-linear classifier (Random Decision Forests) in order to predict the patients grade of facial paresis. In our proposed framework, we make use of highly discriminative features based on the fitting parameters of the Active Appearance Model, Action Units and Landmark distances. We show in our experiments that it is possible to correctly predict the grade of facial paresis in many cases, although the visual appearance is strongly varying. The presented method creates new opportunities to objectively document the patients progress in therapy.