Research Of PET And CT Cross-modality Medical Image Fusion In Radiotherapy

Medical image fusion has become a hotspot in research of application of medical image processing.The technology of medical image fusion has been applied mainly in radiotherapy of tumor to provide precise information in target volume location.PET and CT images are widely used in this field.Firstly,we analyse the principle and structure of both CT scan locating frame and head and neck locating frame which are two important parts in a precise body positioning device for IMRT. Through accurate experiments,an external fiducial marker frame is designed for body and head scan.Our design is made of three organic glass plates looks like a U-shaped structure, including three Z-shaped soft plastic marker tube embedded in every organic glass plate. These marker tubes are hollow which external and internal diameters are 0.6cm and 0.4cm. There are coordinates scale on every organic glass plate. We can get clear markers on PET and CT images by using our design and register PET and CT images with these markers.In addition,by using the external fiducial marker frame designed,rigid medical image registration algorithm is studied in this paper. Firstly,preprocessing PET and CT images by simple segmentation,and then use template matching based on the correlation image detection method to identify the markers' corresponding coordinates,next ,a rigid transform matrix is gotten to register PET and CT images accurately. Error in registration can be greatly reduced by well designing the Z-shaped marker tube in the external fiducial marker frame and also by restricting the relative position between the device and patients. Experimental results show that the external markers registration can be effective and accurate in achieving the PET and CT registration.In order to improve the spatial resolution of PET images, anatomical positioning tumor volume precisely, we propose a new algorithm for PET/CT image fusion by using a variational model based on the wavelet transform after analysis of the characteristics of PET and CT images. Under the general framework of image fusion based on wavelet transform,the first treatment is PET and CT images wavelet decomposition. When choosing image fusion rules, two basic fusion assumptions are proposed in accordance with the PET/CT images,then a variational energy functional model is founded based on these two reasonable assumptions.According to frequency characteristic of wavelet decomposition, we firstly fuse the image in approximate channel by the weighted sum method according to our first assumption,then fuse the image in detail channel by the mean curvature flow method according to our second assumption. The experimental results show that our variational image fusion model can provide more accurate result for target volume location in radiotherapy planning.We study PET and CT cross-modality image fusion in this paper. Each module of our work will be improved combining clinical characteristics of radiotherapy in the future,and preliminary radiotherapy plan will be designed,measuring the results of image fusion and evaluating the plan.