Application Of Auto-context Model In Liver Segmentation From3-D CT Images

China is a country of high incidence of liver disease. Liver surgery is one of themain methods for the treatment of a variety of common benign and malignant liverdisease. Computer-aided planning before liver surgery is an important tache in the op-eration. It plays an important guiding role in the follow-up clinical operation. Liversegmentation from medical images, three-dimensional reconstruction of liver, and liv-er volume measurement are important steps in computer-aided liver surgery planning.Among them, accurate segmentation of liver tissue from medical images is an essen-tial and crucial step for computer-aided liver disease diagnosis and surgical planning.Among the various medical imaging techniques, CT images are often used for thesepurposes, thanks to their higher signal-to-noise ratio and better spatial resolution.However, liver segmentation from CT images is a challenging task, due to thehigh intensity similarity between liver tissue and adjacent organs, the highly varyingshape of the liver, and the presence of severe pathologies. Due to the difculties out-lined above, systems commonly employed in clinical practice rely on manual segmen-tation, which is tedious, time-consuming, and generally not reproducible. Therefore,for clinical application, the fast and accurate3-D CT liver automatic segmentation orinteractive segmentation method is very meaningful. In this paper, we focus on liv-er segmentation from3-D CT images base on Auto-context Model (ACM). The mainwork and contribution of this paper are as follows:This paper gives a review on liver segmentation from3-D CT images. The for-mulation and application for Auto-context Model (ACM) are also explained indetail. · In complicated environment, context information plays an important role in im-age segmentation/labeling. The recently proposed auto-context algorithm is one of the effective context-based methods. However, the original auto-context ap-proach samples the context locations utilizing a fixed radius sequence, which is sensitive to large scale-change of objects. This paper presents a scale invariant auto-context (SIAC) algorithm which is an improved version of the auto-context algorithm. In order to achieve scale-invariance, we try to approximate the opti-mal scale for the image in an iterative way and adopt the corresponding optimal radius sequence for context location sampling, both in training and testing. In each iteration of the proposed SIAC algorithm, we use the current classification map to estimate the image scale, and the corresponding radius sequence is then used for choosing context locations. The algorithm iteratively updates the clas-sification maps, as well as the image scales, until convergence. We demonstrate the SIAC algorithm on several image segmentation/labeling tasks. The result-s demonstrate significant improvement over the original auto-context algorithm when large scale-change of objects exists.· This paper presents a learning-based algorithm for segmenting liver from3D C-T images, by using auto-context model and level-set technique. In the training stage, given a set of abdominal CT training images and their corresponding man-ual liver labels, we first align all the images onto the space of one typical image (randomly selected from the training images), and then extract image and context features for each voxel of all the aligned images. Finally, Auto-context model is performed to train a sequence of classifiers. In the segmentation stage, given a test abdominal CT image, we first align it with the typical image, and then apply the sequence of learned classifiers to compute the classification map (labeling each voxel in the aligned test image). Finally, the classification map will be transformed onto the original test image space and then further transformed into a binary segmentation of liver by a level-set algorithm. The proposed method is evaluated on MICCAI2007liver segmentation challenge datasets. The experi-mental results demonstrate availability of our method.· This paper presents an Auto-context Model (ACM) based automatic liver seg- mentation algorithm, which combines ACM, multi-atlases and mean-shift tech-niques to segment liver from3D CT images. Our algorithm is a learning-based method and can be divided into two stages. At the first stage, i.e. the training stage, ACM is performed to learn a sequence of classifiers in each atlas space (based on each atlas and other aligned atlases). With the use of multiple atlases, multiple sequences of ACM-based classifiers are obtained. At the second stage, i.e. the segmentation stage, the test image will be segmented in each atlas space by applying each sequence of ACM-based classifiers. The final segmentation result will be obtained by fusing segmentation results from all atlas spaces via a multi-classifier fusion technique. Specially, in order to speed up segmentation, given a test image, we first use an improved mean-shift algorithm to perform over-segmentation and then implement the region-based image labeling instead of the original inefficient pixel-based image labeling. The proposed method is evaluated on the datasets of MICCAI2007liver segmentation challenge. The experimental results show that the average volume overlap error and the aver-age surface distance achieved by our method are8.3%and1.5mm, respectively, which are comparable to the results reported in the existing state-of-the-art work on liver segmentation.· This paper proposes a new active contour algorithm, i.e. Hierarchical Contextual Active Contour (HCAC), and apply it to liver segmentation from3D CT images. HCAC is a learning-based method and can be divided into two stages. At the first stage, i.e. the training stage, given a set of abdominal CT training images and the corresponding manual liver labels, we try to establish a mapping between au-tomatic segmentations (in each round) and manual reference segmentations via context features, and obtain a series of self-correcting classifiers. At the second stage, i.e. the segmentation stage, we first use the basic active contour to segment the image and subsequently Contextual Active Contour (CAC), which combines the image information and the current shape model, is iteratively performed to improve the segmentation result. The current shape model is produced by the corresponding self-correcting classifier (the input is the previous automatic seg-mentation result). The proposed method is evaluated on the datasets of MICCAI 2007liver segmentation challenge. The experimental results show that we obtain more and more accurate segmentation results by the iterative steps and satisfying results are obtained after about six iterations.