In this paper, we describe an efficient method for 3D image segmentation. The method uses a PDE model - the so called generalized subjective surface equation which is an equation of advection-diffusion type. The main goal is to develop an efficient and stable numerical method for solving this problem. The numerical solution is based on semi-implicit time discretization and flux-based level set finite volume space discretization. The space discretization is discussed in details and we introduce three possible alternatives of the so called diamond cell finite volume scheme for this type of 3D nonlinear diffusion equation. We test the performance of the method and all its variants introduced in the paper by determining the experimental order of convergence. Finally we show a couple of practical applications of the method.
The purpose of this paper is to develop some effective robust fuzzy c-means methods for segmentation of Brain Medical Images and Dynamic Contrast-Enhanced Breast Magnetic Images (DCE-BMRI). Segmentation is a difficult task and challenging problem in the brain and breast medical images for diagnosing Breast and Brain cancer related diseases before the image goes for treatment plan. This paper presents three new effective fuzzy clustering techniques: Robust KFCM (Kernel Fuzzy C-Means) with spatial information, Effective Robust FCM based Kernel function, Modified fuzzy c-means algorithm with weight Bias Estimation. In experiments, the presented methods are compared with other reported methods. Experimental results on both breast and brain MR images show that the proposed algorithms have better performance than the standard algorithms. Thus, the proposed method is capable of dealing with the intensity in-homogeneities and noised image effectively.
Intensity inhomogeneity or intensity non-uniformity (INU) is an undesired phenomenon that represents the main obstacle for magnetic resonance (MR) image segmentation and registration methods. Various techniques have been proposed to eliminate or compensate the INU, most of which are embedded into clustering algorithms, and they generally have difficulties when INU reaches high amplitudes. This study proposes a multiple stage fuzzy c-means (FCM) clustering based algorithm for the estimation and compensation of INU, by modelling it as a slowly varying additive or multiplicative noise. The slowly varying behaviour of the bias or gain field is assured by a smoothing filter that performs a context dependent averaging, controlled by a morphological criterion. The segmentation is also supported by a prefiltering technique for Gaussian and impulse noise elimination. The experiments using 2-D synthetic phantoms and real MR images indicate that the proposed method provides accurate segmentation. The resulting segmentation and fuzzy membership values can serve as excellent support for 3-D registration and surface reconstruction techniques.
This paper presents a new model to perform a supervised image segmentation task. The proposed model is called segmentation and classification with receptive fields (SCRF) which is based on the concept of receptive fields that analyzes pieces of an image considering not only a pixel or a group of pixels, but also the relationship between them and their neighbors. In order to work with the SCRF model, we propose a new artificial neural network, called I-PyraNet, which is a hybrid implementation of the recently described PyraNet and the nonclassical receptive fields inhibition. Furthermore, the model and the neural network are combined to accomplish a satellite image segmentation task.
The article focuses on the application of the segmentation algorithm based on the numerical solution of the Allen-Cahn non-linear diffusion partial differential equation. This equation is related to the motion of curves by mean curvature. It exhibits several suitable mathematical properties including stable solution profile. This allows the user to follow accurately the position of the segmentation curve by bringing it quickly to the vicinity of the segmented object and by approaching the details of the segmentation curve. The purpose of the article is to indicate how the algorithm parameters are set up and to show how the algorithm behaves when applied to the particular class of medical data. In detail we describe the algorithm parameters influencing the segmentation procedure. The left ventricle volume estimated by the segmentation of scanned slices is evaluated through the cardiac cycle. Consequently, the ejection fraction is evaluated. The described approach allows the user to process cardiac cine MR images in an automated way and represents, therefore, an alternative to other commonly used methods. Based on the physical and mathematical background, the presented algorithm exhibits the stable behavior in the segmentation of MRI test data, it is computationally efficient and allows the user to perform various implementation improvements.
The Self-Organizing Map model considers the possibility of 1D and 3D map topologies. However, 2D maps are by far the most used in practice. Moreover, there is a lack of a theory which studies the relative merits of 1D, 2D and 3D maps. In this paper a theory of this kind is developed, which can be used to assess which topologies are better suited for vector quantization. In addition to this, a broad set of experiments is presented which includes unsupervised clustering with machine learning datasets and color image segmentation. Statistical significance tests show that the 1D maps perform significantly better in many cases, which agrees with the theoretical study. This opens the way for other applications of the less popular variants of the self-organizing map.