Research On Evaluation Method Of Facial Image Aesthetic Outcome After Cleft Lip And Palate Operation

Postoperative patients with cleft lip and palate are sometimes accompanied by scars,lip and nose distortion,etc.Because the evaluation of the patient’s postoperative facial recovery outcome is strongly subjective,after surgery for children with cleft lip and palate,both doctors and patients often complain about Medical disputes occurred due to disagreements on the outcome of recovery.In order to solve this problem,this thesis proposes an objective and feasible method for evaluating the aesthetic outcome of facial images after cleft lip and palate surgery.The evaluation method proposed in this thesis is to first perform lip segmentation and symmetry axis detection on the facial image,then calculate the symmetry of the lips,and then map the symmetry to a score to obtain an aesthetic evaluation.In terms of lip segmentation,this thesis proposes a lip segmentation method based on the improved Mask-RCNN model and fusion strategy.Based on the Mask-RCNN model,a two-way fusion FPN network is used to replace the original one-way FPN network in the model to enhance the effect of image feature fusion;a jump structure is added to the Mask branch network to make the model better at the edge of the segmentation area.At the same time,a fusion strategy method is proposed to solve the problem of scar detection around the lips.In the aspect of detecting symmetry axis,this thesis proposes a face symmetry detection method based on genetic algorithm,which completes individual coding design,fitness function design,genetic operator design,etc.,and searches for the optimal face symmetry axis globally.The lip segmentation result is combined with the symmetry axis,the lip symmetry is calculated through affine transformation and pixels,and finally the aesthetic evaluation is performed based on the lip symmetry.This thesis uses a self-made facial image data set to test the evaluation method.In the lip segmentation comparison experiment,the intersection of the improved Mask-RCNN model’s lip segmentation results is 0.5% and 1.7% higher than the IoU and pixel accuracy respectively.In the symmetry axis detection experiment,the genetic algorithm-based facial symmetry axis detection algorithm proposed in this thesis has advantages in accuracy,detection speed,and stability over other algorithms.In the experiment of the evaluation method,the correlation coefficient between the scoring data of the method proposed in this paper and the scoring data of professionals reached 0.756.It shows that the method proposed in this thesis is feasible and reliable in evaluating the facial aesthetics of patients with cleft lip and palate after surgery.