Research On Kidney Segmentation Using Spiking Cortical Model And Grab Cut

CT as an important tool in medical imaging plays a significant role in the diagnosis and treatment of kidney disease, in which the low-dose CT attracts increasingly common concern due to less harmness to the human body compared with the traditional CT. Diagnosis and treatment of kidney disease often requires to obtain the various parameters related to renal morphology. To measure these parameters accurately, it is necessary to segment the kidney organ from the CT images first. Kidney segmentation has become a hot topic in such fields as renal diagnosis and treatment, surgical planning and surgical navigation. The research on the accurate kidney segmentation has important theoretical and practical significance for the accurate diagnosis and treatment of kidney disease.Grab Cut segmentation method is an effective method for image segmentation, and it has great potential applications to the kidney segmentation. However, the reduction of the radiation dose tends to procude noise, artifacts, adhesion between kidney and other organs in the CT images. These disadvantageous factors will lead to the additional operation of selecting seed points and the inaccurate segmentation of kidney by Grab Cut. To overcome the difficulties, we have proposes a novel method by combining spiking cortical model with Grab Cut, which involves the automatic selection of seed points and the correction of the boundary property term in Grab Cut segmentation method. On one hand, SCM is combined with morphological operators and contour extraction operation to select seed points automatically. On the other hand, the firing image sequences resulting from using SCM are processed in the following ways. The entropy image is firstly computed based on the image patches in each firing image to characterize the image feature. Then the obtained entropy image is utilized to correct the parameter in the boundary term of the Grab Cut energy function. This method will provide an effective means for the accurate kidney segmentation.In this paper, two low-dose abdominal CT image sequences are chosen as the test data. Comarsions between the proposed method and the traditional Grab Cut method are made in terms of objective metrics and human vision. As regards the objective assessment, two metrics, i.e., FOM(Pratt's figure of merit) and DSC(Dice Similarity Coefficient) have been utilized. The FOM and DSC results show that the proposed method is provide with higher segmentation accuaracy than the traditional method. The subjective assessment demonstrates that the method proposed in the paper can overcome the drawbacks of kidney over-segmentation and under-segmentation by Grab Cut, and thus it can realize the effective kidney segmentation.