The Analysis And Evaluation Of Photodynamic Therapy For Port Wine Stains

Photodynamic therapy is a new technology of selective treatment which has showed a good efficacy in a variety of ailments.Because objective and scientific evaluation method for port wine stains is missing,clinicians have to determine the parameters for the clinical treatment in their eyes and experience.It is difficult to control the efficacy accurately,and establish reasonable treatment.So,an objective and quantitative grading evaluation method is very important.This thesis proposes an analysis and evaluation system based on stereo vision technology for port wine stains photodynamic therapy.With the aid of 3D reconstruction,lesion segmentation and lesion registration,the system can realize lesion extraction,lesion grading,and efficacy analysis of PDT objectively and automatically.The major work of this thesis includes:1.A calibration method based on color coded chessboard is developed for the stereo vision system.The red-blue calibration board that can be overlapped with color projection and recognized by different color channel of camera is designed for camera-projector calibration,which provide a possibility to increase the projection area,improve the efficiency and spatial resolution of the camera during treatment,and reduce the distance between the camera and projector optical axis.The proposed method can simultaneously calibrate the camera and projector with less calibration time and fewer number of pictures.2.A 3D reconstruction algorithm using projecting structured light for photodynamic therapy analysis system is designed to reconstruct the spatial information of lesion.Spatial information of lesion is reconstructed as point cloud using the calibrated camera-projector system,in which trigonometry and gray code structure light is applied.Based on the point cloud,the vector of lesion surface is computed by local surface fitting in least square method,which provides the basis for extracting and grading lesion,analyzing efficacy,and correctting brightnession.3.As for the lesion extraction and grading,k-means clustering considering both spatial distance and color disparity in CIEL*a*b* color space is designed and accomplished to realize the point cloud segmentation,which could avoid the impact of environmental noise and matchs the actual block therapy effectivelly.4.A registration algorithm for 3D point cloud is designed to register point cloud of lesion before and after treatment,which makes objective evaluation for efficiency become probable.To verify the proposed evaluation method,a series of experiments is designed and conducted on a color painted head model which is used to simulate the PWS patients.The experimental results and the analysis both on qualitative and quantitative perspective indicate that the system designed in this thesis works well.