Research On The Segmentation Method Of Suspected Malignant Lung Nodules Based On PET/CT Low Dose Scanning

At present, improve the early detection rate of lung cancer through doing regular physical examination of pulmonary for the general population, has been widely recognized. And in order to reduce the amount of radiation, the examination will usually use low dose scanning mode. But this will bring about some problems, such as poor quality of medical image, large amount of image data and low detection rate of lung cancer.In order to solve these problems and improve the quality and efficiency of computer aided diagnosis system for early diagnosis of lung cancer. Considering the suspected malignant pulmonary nodules have obvious imaging in PET images, this paper proposed a segmentation method of suspected malignant pulmonary nodules based on PET/CT low dose scanning to realize the targeted segmentation of suspected malignant pulmonary nodules and lay the foundation for further aided diagnosis. The main research contents are as follows:Firstly, aiming at the poor quality of medical image under low dose, we considering of artifacts and noise have the different characteristics in low dose images, proposes a new low dose image denoising method based on one dimensional diffusion filtering and bilateral filtering. The method is divided into two steps:(1) For the strip artifacts which has direction and vibration in high frequency, this paper proposed a new idea that using one dimensional diffusion filtering with different directions to remove the strip artifacts in different highfrequency.(2) To remove the mottled noise of low dose CT image, we proposed a bilateral filtering based region, which add the regional information to the bilateral filtering algorithm to guide the image denoising. The method was performed on low dose CT images of 25 patients(12 females and 13 males), and achieved a good denoising effect of artifact and noise. The results show that this method can effectively improve the quality of low dose CT images.Secondly, according to the problem that the general method can’t targeted segment suspected malignant pulmonary nodules, a new active contour model automatic segmentation algorithm based on PET/CT is proposed in this paper, which can combined PET image in that malignant nodules have obvious imaging, and CT image in that nodules structure can been obviously show. This method can been divided into three step:(1) Threshold segmentation and regional growth segmentation are combined to achieve lung parenchyma segmentation on CT.(2) Variable template matching is used to coarse segment the suspected malignant lung nodule on PET lung parenchyma which is obtain by registration of CT lung parenchyma.(3) An active contour model is used to accurately segment suspected malignant lung nodule on CT, initial contour is obtained by corresponding the edge of suspected malignant lung nodule in PET to the CT. The results show that the proposed method can be targeted accurately segment the suspected malignant lung nodules.