Iterative Conditional Model Research Articles

A Variable Weight Image Segmentation Algorithm Based on Markov Random Field

Proposed a weighted image segmentation method based on Markov Random Field and context information of the image. On the basis of traditional potential function, the paper introduces the relationship of neighborhood pixels to describe the probability of pixels being partitioned into the same class. We convert image segmentation problem into maximum a posterior for extreme value by Bayes theory. Meanwhile, we introduce the variable weight method to connect feature field and label field. Finally, the iterative conditional model is applied to acquire optimal segmentations. Compared to K-means and traditional MRF algorithm, the proposed method presents effectiveness and robustness in the experiments.

An integrated approach to segmentation and nonrigid registration for application in image-guided pelvic radiotherapy

External beam radiotherapy (EBRT) has become the preferred options for nonsurgical treatment of prostate cancer and cervix cancer. In order to deliver higher doses to cancerous regions within these pelvic structures (i.e. prostate or cervix) while maintaining or lowering the doses to surrounding non-cancerous regions, it is critical to account for setup variation, organ motion, anatomical changes due to treatment and intra-fraction motion. In previous work, manual segmentation of the soft tissues is performed and then images are registered based on the manual segmentation. In this paper, we present an integrated automatic approach to multiple organ segmentation and nonrigid constrained registration, which can achieve these two aims simultaneously. The segmentation and registration steps are both formulated using a Bayesian framework, and they constrain each other using an iterative conditional model strategy. We also propose a new strategy to assess cumulative actual dose for this novel integrated algorithm, in order to both determine whether the intended treatment is being delivered and, potentially, whether or not a plan should be adjusted for future treatment fractions. Quantitative results show that the automatic segmentation produced results that have an accuracy comparable to manual segmentation, while the registration part significantly outperforms both rigid and nonrigid registration. Clinical application and evaluation of dose delivery show the superiority of proposed method to the procedure currently used in clinical practice, i.e. manual segmentation followed by rigid registration.

Mixture model and Markov random field-based remote sensing image unsupervised clustering method

AbstractIn this paper, a novel method for remote sensing image clustering based on mixture model and Markov random field (MRF) is proposed. A remote sensing image can be considered as Gaussian mixture model. The image clustering result corresponding to the image label field is a MRF. So, the image clustering procedure is transformed to a maximum a posterior (MAP) problem by Bayesian theorem. The intensity difference and the spatial distance between the two pixels in the same clique are introduced into the traditional MRF potential function. The iterative conditional model (ICM) is employed to find the solution of MAP. We use the max entropy criterion to choose the optimal clustering number. In the experiments, the method is compared with the traditional MRF clustering method using ICM and simulated annealing (SA). The results show that this method is better than the traditional MRF model both in noise filtering and miss-classification ratio.

Adaptive segmentation of cerebrovascular tree in time-of-flight magnetic resonance angiography

Accurate segmentation of the human vasculature is an important prerequisite for a number of clinical procedures, such as diagnosis, image-guided neurosurgery and pre-surgical planning. In this paper, an improved statistical approach to extracting whole cerebrovascular tree in time-of-flight magnetic resonance angiography is proposed. Firstly, in order to get a more accurate segmentation result, a localized observation model is proposed instead of defining the observation model over the entire dataset. Secondly, for the binary segmentation, an improved Iterative Conditional Model (ICM) algorithm is presented to accelerate the segmentation process. The experimental results showed that the proposed algorithm can obtain more satisfactory segmentation results and save more processing time than conventional approaches, simultaneously.

Edge detection using markov random field models—optimization and parameter estimation by mean field annealing

AbstractAn edge‐detection method based on Markov random field and statistical estimation is proposed. Compound Gauss‐Markov random fields are used as the image model, and the edge‐detection problem was formulated as the maximum a posteriori probability (MAP) estimation problem which can determine the estimated value of line process from the observed value of intensity process. the simulated annealing (SA) method and iterative conditional model (ICM) method are well known as relaxation methods for such types of problems. However, a large amount of calculations is needed in the SA method and convergence to local optimum solution is a major problem in the ICM method.This paper determines the MAP estimation solution using the method of mean field annealing (MFA), which has shown good results for the combinatorial optimization problem. Moreover, since in this method the shape and continuity of the detected edge changes with the model parameters of compound Gauss‐Markov random field, parameter estimation is an important problem.In the previous research, those parameter estimation methods have been studied where parameters determined from experience have been used or parameters are estimated using the edges traced by a human being. This paper proposes the method of automatic parameter estimation from the observed image based on the maximum likelihood (ML) criterion. Moreover, the gradient descent method and expectation maximization (EM) algorithm are used to calculate the actual ML solution.