Facial paralysis affects both mental and physical health of patients. Evaluation of facial paralysis in clinical practice is normally based on the static facial asymmetry at the maximal movement state and the dynamic change of facial muscle movement. However, most existing research only considers one of these two aspects when evaluating the degree of facial paralysis. This will result in an incomplete utilization of the diagnosis information leading to a low evaluation accuracy, or even misjudgment. In this paper, a novel method is presented for evaluating the degree of facial paralysis considering both static facial asymmetry and dynamic transformation factors. A quantitative approach of static facial asymmetry based on local mirror asymmetry is proposed. The method compares the differences of the corresponding local regions between both sides of the face. This makes it effective analyzing the left and right side asymmetry for abnormal faces. A quantitative evaluation of static facial asymmetry is achieved through three steps: localization of local facial regions, extraction of asymmetry features and quantification of bilateral facial asymmetry. Once the static facial asymmetry is quantified, its dynamic counterparts can be calculated using the speed of changings in different regions caused by facial muscle movement. Then we combine the static and dynamic quantification results to evaluate the degree of facial nerve paralysis. The efficiency and effectiveness of the proposed method have been tested using our facial paralysis database with 62 patients. The experiments show that our method produced encouraging performance compared with ground truth.
|Journal||Multimedia Tools and Applications|
|Early online date||5 Jun 2015|
|Publication status||Published - Oct 2016|