Diagnosis Support of Patients with Facial Paresis

Computers are used more and more in medical applications. Well{known examples are the analysis of radiographs or MR images 12, 9]. A special eld is the analysis of human faces and facial features. Already realized applications exist e.g. for face recognition 1, 2] or face analysis 13, 6]. With the presented system we consider the problem of diagnosis support of patients with facial paresis. Approximately 350 patients per year are registered in the Department of Otorhino{Larygology of the University of Erlangen with new occurrences of this type of paresis (cf. 17] as an overview). The current way to diagnose the paresis is a subjective judgement of the functionality of the face muscles by a physician. The patient performs mimic exercises such as closing the eyes or showing the teeth, while he is observed by the physician. The subjective observations of the physician are then graded by means of two medical indexing systems 8, 14]. One part of the rehabilitation of the patients is to perform speciic mimic exercises with the face's musculature. The success of the therapy is also observed by a physician or educated clinic personal. For every observation the patient has to travel to the clinics. Furthermore, educated personal is needed for the diagnosis and rehabilitation observation of the paresis. The applications of our system are on the one hand diagnosis support in the clinics. We want to improve the subjective judgments of a physician by objective measurements and numerical features from the face. On the other hand the supervision of the rehabilitation process of the patient will be enhanced by placing the system to the patient's home. The patient can use the system in a convenient environment and does not have to travel to the clinics. The patient does not have to pay for the more convenient observation environment by wearing any artiicial markers inside the face and he is allowed to move in front of the system's camera. We present two diierent approaches for the analysis of facial features which will measure the face during the performance of mimic exercises. The result is a value for every picture with the level of asymmetry of the eye and mouth region which can be used to classify the facial paresis. The asymmetry of the face can be used as we handle just patients with paresis in one half of the face. This class of patients is the most …