Research On Tissue Segmentation Algorithms In Medical Images Using Level Set

With the development of medical imaging technology and computer technology,multi-model imaging techniques(such as Computer Tomography(CT),Positron Emission Tomography(PET),Magnetic Resonance Imaging(MR),Ultrasound Image,etc.)for clinical doctors to diagnosis,treatment of disease,surgery and postoperative evaluation plays and effective assistant function.Generally speaking,early diagnosis of disease can effectively detect diseases in hidden stage,improve disease cure rate,and reduce the treatment cost as much as possible for patients.Multi-modality imaging techniques can completely delineate all the organs and tissues in the body and other pathological information,therefore,how to obtain patients’ various quantified and quantized data accurately and comprehensively becomes the key problem of disease diagnosis.Medical image segmentation techniques can be employed for clinicians to extract regions of interest(such as organs or tissues of human body or lesions,etc.)with quantitative evaluation as well as 3D reconstruction and visualization of regions of interest.This dissertation focuses on the three types of diseeases which cause serious damage to human body health(Alzheimer’s disease,chronic kidney disease and lung cancer)at present,and proposed automated and accurate segmentation methods with the aim of satisfied the clinical diagnosis of surgical planning,early prediction of diseases,even the postoperative recovery treatment.Level set segmentation method is one of the most popular segmentation algorithms in medical image segmentation field.The main idea of level set approach is expressing a closed contour as a set of high dimensional curved contour even points,with the action of external force and internal force,zero level set is evolution by level sets.When the energy function reaches a minimization,level set evolution will stop at the target contour.In order to propose automatic and efficient segmentation algorithms to satisfy the clinical diagnosis of the surgery planning,and early disease prediction,or even postoperative recovery,based on multi-modality imaging characteristics,we focus on level set formulation;we proposed fully automatic and accurate medical image tissue segmentation methods,which obtained satisfied results.The thesis consists of the following:1)In this dissertation,we propose a new variational method for performing label fusion that takes into account the image information and regularity of the region of interest(ROI).We also propose to use labels from multiple warping methods,thereby further leveraging upon the complementary advantages of different registration methods.In our approach,each label derived from different atlases and registration methods is represented by a level set function whose zero level contours encloses the labeled region.Label fusion is achieved by seeking an optimal level set function that minimizes energy functional with three terms:label fusion term,image based term,and regularization term.The curve evolution derived from the energy minimization problem is impacted by the three terms simultaneously to achieve optimal label fusion subject to image-based and shape regularity constraints.2)In this dissertation,an improved level set method is proposed for segmentation of Renal Parenchymal Area from ultrasound images based on a 2-step level set method.The proposed method is novel in three aspects.A modified distance regularized level set evolution model is adopted to extract the boundary of kidney,followed by region-scalable fitting model to extract the collecting system within kidney,and measure Renal Parenchymal Area by subtracting the area of the collecting system from the gross kidney area.The proposed method proves to be valid and reliable through experiments on 10 selected ultrasound images despite intra-observer reliability in measurements of Renal Parenchymal Area with obtaining a highest Dice of 0.985 with Dice Score higher than 0.9 for all cases.3)Lymph node detection is challenging due to the low contrast between lymph nodes as well as surrounding soft tissues and the variation in nodal size and shape.In this dissertation,we propose several novel ideas which are combined into a system to operate on positron emission tomography-computed tomography(PET-CT)images to detect abnormal thoracic nodes.First,our previous Automatic Anatomy Recognition(AAR)approach is modified where lymph node zones predominantly following International Association for the Study of Lung Cancer(IASLC)specifications are modeled a subjects arranged in a hierarchy along with key anatomic anchor objects.This fuzzy anatomy model built from diagnostic CT images is then deployed on PET-CT images for automatically recognizing the zones.A novel globular filter(g-filter)to detect blob-like objects over a specified range of sizes is designed to detect the most likely locations and sizes of diseased nodes.Abnormal nodes within each automatically localized zone are subsequently detected via combined use of different items of information at various scales:lymph node zone model poses found at recognition indicating the geographic layout at the global level of node clusters,g-filter response which hones in on and carefully selects node-like globular objects at the node level,and CT and PET gray value but within only the most plausible nodal regions for node presence at the voxel level.The models are built from 25 diagnostic CT scans and refined for an object hierarchy based on a separate set of 20 diagnostic CT scans.Finally,the results are used as the input value of the level set function,and then the segmentation results of the lymph nodes are determined.In this dissertation,the experimental data is from patients with lung cancer in the hospital of Philadelphia.The experimental results show that the segmentation efficiency is high,and can meet the requirements of clinical diagnosis for automatic segmentation of lymph nodes.