Evaluation the Deformity of Cleft Nose based on Statistical Analysis and Neural Network

摘要

An objective quantitative evaluation system of the appearance of the nostril is very important to improve the outcome of surgery. The feature factors are abstracted from the digital image based on the consideration of symmetry. We measure the angles and distance considering the symmetry aspect. Two groups of raters assign a score to the deformity of the cleft lip nose photos. Both a statistical method and neural network are applied to explain how to assess the deformity. In order to obtain more reliable data, this paper proposes the use of Kendall's coefficient of concordance (W) to identify the consistency between several raters. In addition, we compare the linear regression and Neural Network (NN) methods in testing the data to determine the consistency. Then we build linear regression equations describing the relationship between the selected factors and elementary score, in order to obtain more reliable reference data. Finally, we use the NN to predict the evaluation score. The NN can perform better than the linear regression method under certain conditions. The proposed method can give an objective evaluation to help surgeons after plastic surgery and can be directly applied to evaluate possible further treatment.