A comparative study of Image Region-Based Segmentation Algorithms

Synthetic magnetic resonance imaging (SyMRI) allows to obtain different weighted-images using the multiple-dynamic multiple-echo sequence lasting 6min. The aim is to compare quantitatively and qualitatively synthetic- and conventional MRI in patients with multiple sclerosis (MS) and controls assessing the contrast (C), the signal to noise ratio (SNR), and the contrast to noise ratio (CNR). We evaluated the lesion count and lesion-to-white matter contrast ([Formula: see text] in the MS patients. 51 patients underwent synthetic- and conventional MRI. Qualitative analysis was evaluated by assigning scores to all synthetic- and conventional MRI sequences by two neuroradiologists. Lesions were counted in MS patients both in the conventional- and synthetic T2-FLAIR. Regions of interest were placed in the cerebrospinal fluid, in the white- and grey matter. For the sequences were evaluated: C, CNR, and SNR. Synthetic T2-FLAIR images are qualitatively inferior. C and CNR were significantly higher in synthetic T1W and T2W images compared to conventional images, but not for T2-FLAIR. The SNR value was always lower in synthetic images than in conventional ones. SyMRI can be used in clinical practice because it has a similar diagnostic accuracy which reduces the scanning time compared to the conventional one. However, synthetic T2-FLAIR images need to be improved.

The reliability of relaxation time measures in synthetic magnetic resonance images (MRIs) of homemade phantoms were validated, and the diagnostic suitability of synthetic imaging was compared to that of conventional MRIs for detecting ischemic lesions. Phantoms filled with aqueous cupric-sulfate (CuSO4) were designed to mimic spin-lattice (T1) and spin-spin (T2) relaxation properties and were used to compare their accuracies and stabilities between synthetic and conventional scans of various brain tissues. To validate the accuracy of synthetic imaging in ischemic stroke diagnoses, the synthetic and clinical scans of 18 patients with ischemic stroke were compared, and the quantitative contrast-to-noise ratios (CNRs) were measured, using the Friedman test to determine significance in differences. Results using the phantoms showed no significant differences in the interday and intersession synthetic quantitative T1 and T2 values. However, between synthetic and referenced T1 and T2 values, differences were larger for longer relaxation times, showing that image intensities in synthetic scans are relatively inaccurate in the cerebrospinal fluid (CSF). Similarly, CNRs in CSF regions of stroke patients were significantly different on synthetic T2-weighted and T2-fluid-attenuated inversion recovery images. In contrast, differences in stroke lesions were insignificant between the two. Therefore, interday and intersession synthetic T1 and T2 values are highly reliable, and discrepancies in synthetic T1 and T2 relaxation times and image contrasts in CSF regions do not affect stroke lesion diagnoses. Additionally, quantitative relaxation times from synthetic images allow better estimations of ischemic stroke onset time, consequently increasing confidence in synthetic MRIs as diagnostic tools for ischemic stroke.

In present times, automatic computer interpretation of medical images is of great importance for medical diagnosis. Having previously been an obstacle for medical science, early stage diagnoses together with accurately calculating the size and level of disease are very important elements of today's treatment process. The treatment and diagnosis of brain tumors depend on successful processing and evaluation of magnetic resonance (MR) brain images. Image segmentation is the most intractable process, because of the complicated nature of the MR brain images. In processing and analyzing MR images, there is unsuccessful automatic systems or successful semiautomatic system, which is still required large scale experting involvement and assessment. Within this paper the aim is to propose a system that will automatically extract the location, volume and borders of a brain tumor from patient MRI brain images. To achieve this goal, an automated algorithm has been developed, based upon and derived from a successful semi-automatic hybrid algorithm, for automation of the tumor diagnosis process. Patients' MR brain images from varying stages of the diagnosis and treatment were used for testing and validating the automatic algorithm under development. Based upon the achieved results of this algorithm and method, an automatic, sensitive MR brain computer interpretation can be proposed.

The segmentation of magnetic resonance (MR) brain images is an important problem in medical imaging. Accurate segmentation of MR brain images allows a detail study of 3D brain tissue anatomy. It is also of great interest in the study of many brain disorders, where accurate volumetric measurement of the disorders is often required. In view of the importance of the task, much effort has been spent on finding accurate and efficient algorithms for the MRI segmentation problem. This chapter attempts to give the readers an overview of the MR brain segmentation problem, the various image artifacts that are often encountered, and describe some of the current approaches in this area, as well as our own work. To facilitate discussion, we broadly divide current MR brain image segmentation algorithms into three categories: classification-based, region-based, and contour-based methods, and discuss the merits and limitations of these approaches. Following a review of existing methods, we describe our approach for MR brain image segmentation in detail. Our approach is based on a clustering-for-classification framework, using a novel variant of the fuzzy c-means algorithm. We show that by incorporating two key ideas into the clustering algorithm, we are able to take into account the local spatial context, to compensate for the intensity nonuniformity artifact and the partial volume averaging artifact, and to reduce the influence of image noise, during the segmentation process. Extensive experiment results on both simulated and real MR brain images are given to illustrate the effectiveness and robustness of our approach. We conclude this chapter by pointing to some possible future directions in this area.

Brain magnetic resonance (MR) image segmentation is pivotal for quantitative brain analyses, in which statistical models are most commonly used. However, in spite of its computational effectiveness, these models are less capable of handling the intensity non-uniformity (INU) and partial volume effect (PVE), and hence may produce less accurate results. In this paper, a novel brain MR image segmentation algorithm is proposed. To address the INU and PVE, voxel values in each small volume are characterized by a local variational Bayes (LVB) model, which is inferred by the niche differential evolution (NDE) technique to avoid local optima. A probabilistic brain atlas is constructed for each image to incorporate the anatomical prior into the segmentation process. The proposed NDE-LVB algorithm has been compared to the variational expectation-maximization based and genetic algorithm based segmentation algorithms and the segmentation routine in the widely used statistical parametric mapping package on both synthetic and clinical brain MR images. Our results suggest that the NDE-LVB algorithm can differentiate major brain tissue types more effectively and produce more accurate segmentation results.KeywordsImage segmentationMagnetic resonance imaging (MRI)Gaussian mixture modelVariational Bayes inferenceNiche differential evolution

Segmentation of brain magnetic resonance (MR) images is a vital requisite in health care applications. Spatial Fuzzy C-means (SFCM) clustering method segments the brain MR images degraded with noise and artifacts. Due to the iterative computation of the clustering parameters, the run time is high. To address this issue, we suggest a fast spatial FCM (FSFCM) clustering technique for brain MR image segmentation. The basic idea is that, the input image is processed through morphological reconstruction and local spatial similarity factor calculation prior to the clustering process. The morphological reconstruction operation helps in eliminating the noise, while inclusion of the local spatial similarity factor preserves the structural details. The proposed technique is relatively faster than SFCM because the local spatial similarity factor is to be computed only once prior to computation of the clustering parameters. The proposed method is investigated with various volumes of synthetic and clinical brain MR images. The outcomes of the experiment demonstrate better results in comparison to the state-of-the-art methods.

The hidden Markov random field (HMRF) model is an efficient method for image segmentation. However, it may not accurately segment a poorly defined image like Magnetic Resonance (MR) brain image. The HMRF model uses an initial estimation for tissue segmentation of MR brain image. The initial estimation of an input image has a large role to play to obtain a robust HMRF model. In general, a hard clustering method is used for initial estimation. However, it is not an efficient technique for the segmentation of a complex and irregular shape image like MR brain image. The method of clustering acts as a preprocessing step in improving the performance of HMRF. It aims at producing a set of concise data in the segmentation of brain tissues. In this paper, we propose a new approach to HMRF model, which incorporates the fuzzy C-means (FCM) clustering for the initial estimation of the brain image, which is not found in the literature. This results in efficient segmentation of brain tissues. The results are compared with the spatial FCM clustering algorithm and HMRF model with K-means clustering, using different segmentation evaluation indices. It is seen that the results are better than the other methods.

In general, synthetic aperture radar (SAR) imaging and image processing are two sequential steps in SAR image processing. Due to the large size of SAR images, most image processing algorithms require image segmentation before processing. However, the existence of speckle noise in SAR images, as well as poor contrast and the uneven distribution of gray values in the same target, make SAR images difficult to segment. In order to facilitate the subsequent processing of SAR images, this paper proposes a new method that combines the back-projection algorithm (BPA) and a first-order gradient operator to enhance the edges of SAR images to overcome image segmentation problems. For complex-valued signals, the gradient operator was applied directly to the imaging process. The experimental results of simulated images and real images validate our proposed method. For the simulated scene, the supervised image segmentation evaluation indexes of our method have more than 1.18%, 11.2% and 11.72% improvement on probabilistic Rand index (PRI), variability index (VI), and global consistency error (GCE). The proposed imaging method will make SAR image segmentation and related applications easier.

An automated hybrid approach using clustering and nature inspired optimization technique for improved tumor and tissue segmentation in magnetic resonance brain images

Pretreatment synthetic MRI features for triple-negative breast cancer

A simple yet accurate method for segmenting magnetic resonance (MR) brain images has been implemented. The semi-automatic technique carries out region growing in all three dimensions guided by initial seed points. Seed voxels may be specified interactively with a mouse or automatically through the selection of intensity thresholds. The 3-D seeded region growing (3-D SRG) algorithm offers significant advantages for MR brain segmentation, particularly in terms of speed and flexibility.

Diffusion coupling model based on neural oscillation synchronization theory, which can extract different targets with different gray levels from one, has been an important way for realizing image segmentation. This paper describes the diffusion coupling model and its algorithm and process for image segmentation. Simulation programs for supplementing this algorithm were compiled based on MATLAB platform. Experimental results on synthetic images, infrared images and brain magnetic resonance images show that, by setting reasonable algorithm parameters, the target image extraction and segmentation of different elements can be achieved. This image segmentation algorithm exhibits wide applications in medical image diagnosis, satellite image processing, traffic control, biometrics and others.

In this article, we describe the development and validation of an automatic algorithm to segment brain from extracranial tissues, and to classify intracranial tissues as cerebrospinal fluid (CSF), gray matter (GM), white matter (WM) or pathology. T1 weighted spin echo, dual echo fast spin echo (T2 weighted and proton density (PD) weighted images) and fast Fluid Attenuated Inversion Recovery (FLAIR) magnetic resonance (MR) images were acquired in 100 normal patients and 9 multiple sclerosis (MS) patients. One of the normal studies had synthesized MS-like lesions superimposed. This allowed precise measurement of the accuracy of the classification. The 9 MS patients were imaged twice in one week. The algorithm was applied to these data sets to measure reproducibility. The accuracy was measured based on the synthetic lesion images, where the true voxel class was known. Ninety-six percent of normal intradural tissue voxels (GM, WM, and CSF) were labeled correctly, and 94% of pathological tissues were labeled correctly. A low coefficient of variation (COV) was found (mean, 4.1%) for measurement of brain tissues and pathology when comparing MRI scans on the 9 patients. A totally automatic segmentation algorithm has been described which accurately and reproducibly segments and classifies intradural tissues based on both synthetic and actual images.

Segmentation is the process of extracting structures within the images. The purpose is to simplify the representation of the image into something meaningful and easier to analyse. A magnetic resonance (MR) brain image can be represented as three main tissues, e.g. cerebrospinal fluid (CSF), grey matter and white matter. Although various segmentation methods have been developed, such images are generally segmented by modelling the intensity histogram by using a Gaussian Mixture Model (GMM). However, the standard use of 1D histogram sometimes fails to find the mean for Gaussians. We hence solved this by including gradient information in the 2D intensity and intensity gradient histogram. We applied our methods on real data of 2D MR brain images. We then compared the methods with the previous published method of Petrovic et al. on their dataset, as well as on our larger datasets extracted from the same database of 3D MR brain mages, where the ground-truth annotations are available. This shows that our method performs better than the previous method.

PurposeWe aimed to validate the estimation of the brain parenchymal fraction (BPF) in patients with multiple sclerosis (MS) using synthetic magnetic resonance imaging (SyMRI) by comparison with software tools of the FMRIB Software Library (FSL). In addition to a cross-sectional method comparison, longitudinal volume changes were assessed to further elucidate the suitability of SyMRI for quantification of disease-specific changes.MethodsMRI data from 216 patients with MS and 28 control participants were included for volume estimation by SyMRI and FSL-SIENAX. Moreover, longitudinal data from 35 patients with MS were used to compare registration-based percentage brain volume changes estimated using FSL-SIENA to difference-based calculations of volume changes using SyMRI.ResultsWe observed strong correlations of estimated brain volumes between the two methods. While SyMRI overestimated grey matter and BPF compared to FSL-SIENAX, indicating a systematic bias, there was excellent agreement according to intra-class correlation coefficients for grey matter and good agreement for BPF and white matter. Bland–Altman plots suggested that the inter-method differences in BPF were smaller in patients with brain atrophy compared to those without atrophy. Longitudinal analyses revealed a tendency for higher atrophy rates for SyMRI than for SIENA, but SyMRI had a robust correlation and a good agreement with SIENA.ConclusionIn summary, BPF based on data from SyMRI and FSL-SIENAX is not directly transferable because an overestimation and higher variability of SyMRI values were observed. However, the consistency and correlations between the two methods were satisfactory, and SyMRI was suitable to quantify disease-specific atrophy in MS.