Texture Based Segmentation using Statistical Properties for Mammographic Images

Unsupervised texture-based image segmentation through pattern discovery

Image segmentation is one of the pivotal steps in image processing due to its enormous application potential in medical image analysis, data mining, and pattern recognition. In fact, image segmentation is the process of splitting an image into multiple parts in order to provide detailed information on different aspects of the image. Traditional image segmentation techniques suffer from local minima and premature convergence issues when exploring complex search spaces. Additionally, these techniques also take considerable runtime to find the optimal pixels as the threshold levels are increased. Therefore, in order to overcome the computational overhead and convergence problems of the multilevel thresholding process, a robust optimizer, namely the Levy flight and Chaos theory-based Gravitational Search Algorithm (LCGSA), is employed to perform the segmentation of the COVID-19 chest CT scan images. In LCGSA, exploration is carried out by Levy flight, while chaotic maps guarantee the exploitation of the search space. Meanwhile, Kapur’s entropy method is utilized for segmenting the image into various regions based on the pixel intensity values. To investigate the segmentation performance of ten chaotic versions of LCGSA, firstly, several benchmark images from the USC-SIPI database are considered for the numerical analysis. Secondly, the applicability of LCGSA for solving real-world image processing problems is examined by using various COVID-19 chest CT scan imaging datasets from the Kaggle database. Further, an ablation study is carried out on different chest CT scan images by considering ground truth images. Moreover, various qualitative and quantitative metrics are used for the performance evaluation. The overall analysis of the experimental results indicated the efficient performance of LCGSA over other peer algorithms in terms of taking less computational time and providing optimal values for image quality metrics.

Image segmentation is one of the most attractive problems in image processing. In image segmentation how to extract useful features from image has become crucial. However, color feature or texture feature, which are both wildly used features, could not process segmentation problem alone very well, especially when images are complex. We adopt a rough-fuzzy set approach, which can properly process high dimensionality, for image segmentation considering both color and texture features. This approach firstly constructs a structure named fuzzy data cube, whose attributes are composed of the fuzzy sets associated with image features. The fuzzy data cube, which can be two-dimension or high-dimension, is as the basic data structure in this method. A definition of the membership function of similarity relation based rough-fuzzy set is introduced as well as the definition of dependency function to evaluate the importance of an attribute for image segmentation. Then we used the rough-fuzzy set to discover the similarity set in fuzzy data cube to obtain the segmentation result. Experiments on mosaic and natural images are presented to demonstrate the effectiveness of the proposed method.

Image segmentation is the method of partitioning an image into a group of pixels that are homogenous in some manner. The homogeneity dependents on some attributes like intensity, color etc. Segmentation being a pre-processing step in image processing have been used in the number of applications like identification of objects to medical images, satellite images and much more. The taxonomy of an image segmentation methods collectively can be divided among two categories Traditional methods and Soft Computing (SC) methods. Unlike Traditional methods, SC methods have the ability to simulate human thinking and are flexible to work with their ownership function, have been predominantly applied to the task of image segmentation. SC techniques are tolerant of partial truth, imprecision, uncertainty, and approximations. Soft Computing approaches also having advantages of providing cost-effective, high performance and steadfast solutions. In this survey paper, our emphasis is on core SC approaches like Fuzzy logic, Artificial Neural Network, and Genetic Algorithm used for image segmentation. The contribution lies in the fact to present this paper to the researchers that explore state-of-the-art elaboration of almost all dimensions associated with the image segmentation. The idea is to encapsulate various aspects like emerging topics, methods, evaluation parameters, the problem associated with different type of images, databases, segmentation applications, and other resources so that, it could be advantageous for researchers to make effort in developing new methods for segmentation. The paper accomplishes with findings and concluding remarks.

In this paper, a texture-based segmentation approach using wavelet packets, co-occurrence matrices and normalised modified histogram thresholding is discussed and developed. Background and objects in remotely sensed light detection and ranging (LIDAR) data are successfully partitioned into rivers, fields and residential areas using the developed algorithms. The issue of wavelet packet decomposition level is addressed by analysing the sub-images' energy and entropy. The standard deviation of the modified histogram, which is derived from the main diagonal of the sub-image's co-occurrence matrix, is used as a measure to evaluate the sub-images in terms of thresholdability. Finally, the segmentation results are presented.

This paper presents an unsupervised texture-based segmentation algorithm which uses reduced contourlet transform sub-bands and mean shift clustering, to analysis the texture information of high resolution SAR images. One step and criteria is proposed to reduce the sub-bands and other's is presented to decrease the number of dimension of the feature space. The mean shift clustering method is used to obtain the number of texture regions and the centre of the label class. Group the pixels into corresponding texture region by their simple distance to the class centre pixel. Experiments on a mixture of Brodatz texture and SAR images show the proposed algorithm of using contourlet transform and mean shift clustering gives satisfactory results.

Conventional unsupervised image segmentation methods use color and geometric information and apply clustering algorithms over pixels. They preserve object boundaries well but often suffer from over-segmentation due to noise and artifacts in the images. In this paper, we contribute on a preprocessing step for image smoothing, which alleviates the burden of conventional unsupervised image segmentation and enhance their performance. Our approach relies on a convolutional autoencoder (CAE) with the total variation loss (TVL) for unsupervised learning. We show that, after our CAE-TVL preprocessing step, the over-segmentation effect is significantly reduced using the same unsupervised image segmentation methods. We evaluate our approach using the BSDS500 image segmentation benchmark dataset and show the performance enhancement introduced by our approach in terms of both increased segmentation accuracy and reduced computation time. We examine the robustness of the trained CAE and show that it is directly applicable to other natural scene images.

Medical image processing and analysis operations, particularly segmentation, can benefit a great deal from prior information encoded to capture variations over a population in form, shape, anatomic layout, and image appearance of objects. Model/atlas-based methods are extant in medical image segmentation. Although multi-atlas/ multi-model methods have shown improved accuracy for image segmentation, if the atlases/models do not cover representatively the distinct groups, then the methods may not be generalizable to new populations. In a previous study, we have given an answer to address the following problem at image level: How many models/ atlases are needed for optimally encoding prior information to address the differing body habitus factor in a population? However, the number of models for different objects may be different, and at the image level, it may not be possible to infer the number of models needed for each object. So, the modified question to which we are now seeking an answer to in this paper is: How many models/ atlases are needed for optimally encoding prior information to address the differing body habitus factor for each object in a body region? To answer this question, we modified our method in the previous study for seeking the optimum grouping for a given population of images but focusing on the individual objects. We present our results on head and neck computed tomography (CT) scans of 298 patients.

Medical image computing has become one of the most challenging fields in medical informatics. In image-based diagnostics of the future software assistance will become more and more important, and image analysis systems integrating advanced image computing methods are needed to extract quantitative image parameters to characterize the state and changes of image structures of interest (e.g. tumors, organs, vessels, bones etc.) in a reproducible and objective way. Furthermore, in the field of software-assisted and navigated surgery medical image computing methods play a key role and have opened up new perspectives for patient treatment. However, further developments are needed to increase the grade of automation, accuracy, reproducibility and robustness. Moreover, the systems developed have to be integrated into the clinical workflow. For the development of advanced image computing systems methods of different scientific fields have to be adapted and used in combination. The principal methodologies in medical image computing are the following: image segmentation, image registration, image analysis for quantification and computer assisted image interpretation, modeling and simulation as well as visualization and virtual reality. Especially, model-based image computing techniques open up new perspectives for prediction of organ changes and risk analysis of patients and will gain importance in diagnostic and therapy of the future. From a methodical point of view the authors identify the following future trends and perspectives in medical image computing: development of optimized application-specific systems and integration into the clinical workflow, enhanced computational models for image analysis and virtual reality training systems, integration of different image computing methods, further integration of multimodal image data and biosignals and advanced methods for 4D medical image computing. The development of image analysis systems for diagnostic support or operation planning is a complex interdisciplinary process. Image computing methods enable new insights into the patient's image data and have the future potential to improve medical diagnostics and patient treatment.

Improving the runtime of MRF based method for MRI brain segmentation

The authors present results and the latest advancement in their long-term study on using image processing and data mining methods in medical image analysis in general, and in clinical diagnostics of coronary artery disease in particular. Since the evaluation of modern medical images is often difficult and time-consuming, authors integrate advanced analytical and decision support tools in diagnostic process. Partial diagnostic results, frequently obtained from tests with substantial imperfections, can be thus integrated in ultimate diagnostic conclusion about the probability of disease for a given patient. Authors study various topics, such as improving the predictive power of clinical tests by utilizing pre-test and post-test probabilities, texture representation, multi-resolution feature extraction, feature construction and data mining algorithms that significantly outperform the medical practice. During their long-term study (1995-2011) authors achieved, among other minor results, two really significant milestones. The first was achieved by using machine learning to significantly increase post-test diagnostic probabilities with respect to expert physicians. The second, even more significant result utilizes various advanced data analysis techniques, such as automatic multi-resolution image parameterization combined with feature extraction and machine learning methods to significantly improve on all aspects of diagnostic performance. With the proposed approach clinical results are significantly as well as fully automatically, improved throughout the study. Overall, the most significant result of the work is an improvement in the diagnostic power of the whole diagnostic process. The approach supports, but does not replace, physicians’ diagnostic process, and can assist in decisions on the cost-effectiveness of diagnostic tests.

With the advent of new technologies in the field of medicine, there is rising awareness of biomechanisms, and we are better able to treat ailments than we could earlier. Deep learning has helped a lot in this endeavor. This paper deals with the application of deep learning in brain tumor segmentation. Brain tumors are difficult to segment automatically given the high variability in the shapes and sizes. We propose a novel yet simple fully convolutional network (FCN) which results in competitive performance and faster runtime than state-of-the-art model. Using the database provided for the Brain Tumor Segmentation (BraTS) challenge by the Medical Image Computing and Computer Assisted Intervention (MICCAI) society, we are able to achieve dice scores of 0.83 in the whole tumor region, 0.75 in the core tumor region and 0.72 in the enhancing tumor region, while our method is about 18 times faster than the state-of-the-art.

Image segmentation is a very crucial step in medical image analysis which is the first and the most important task in many clinical interventions. The authors propose in this paper to apply the variable neighborhood search (VNS) metaheuristic on the problem of brain magnetic resonance images (MRI) segmentation. In fact, by reviewing the literature, they notice that when the number of classes increases the computational time of the exhaustive methods grows exponentially with the number of required classes. That's why they exploit the VNS algorithm to optimize two maximizing thresholding functions which are the between-class variance (the Otsu's function) and the entropy thresholding (the Kapur's function). Thus, two versions of the VNS metaheuristic are respectively obtained: the VNS-Otsu and the VNS-Kapur. These two novel proposed thresholding methods are tested on a set of benchmark brain MRI to show their robustness and proficiency.

The prefrontal cortex is commonly associated with cognitive capacities related to human uniqueness: purposeful actions towards higher-level goals, complex social information processing, introspection, and language. Comparative investigations of the prefrontal cortex may thus shed more light on the neural underpinnings of what makes us human. Using histological data from 19 anthropoid primate species (6 apes including humans and 13 monkeys), we investigate cross-species relative size changes along the anterior (prefrontal) and posterior (motor) axes of the cytoarchitectonically defined frontal lobe in both hemispheres. Results reveal different scaling coefficients in the left versus right prefrontal hemisphere, suggest that the primary factor underlying the evolution of primate brain architecture is left hemispheric prefrontal hyperscaling, and indicate that humans are the extreme of a left prefrontal ape specialization in relative white to grey matter volume. These results demonstrate a neural adaptive shift distinguishing the ape from the monkey radiation possibly related to a cognitive grade shift between (great) apes and other primates.

We propose a novel approach to investigate and implement unsupervised image content understanding and segmentation of color industrial images like medical imaging, forensic imaging, security and surveillance imaging, biotechnical imaging, biometrics, mineral and mining imaging, material science imaging, and many more. In this particular work, our focus will be on medical images only. The aim is to develop a computer aided diagnosis (CAD) system based on a newly developed Multidimensional Spatially Variant Finite Mixture Model (MSVFMM) using Markov Random Fields (MRF) Model. Unsupervised means automatic discovery of classes or clusters in images rather than generating the class or cluster descriptions from training image sets. The aim of this work is to produce precise segmentation of color medical images on the basis of subtle color and texture variation. Finer segmentation of images has tremendous potential in medical imaging where subtle information related to color and texture is required to analyze the image accurately. In this particular work, we have used CIE-Luv and Daubechies wavelet transforms as color and texture descriptors respectively. Using the combined effect of a CIE-Luv color model and Daubechies transforms, we can segment color medical images precisely in a meaningful manner. The evaluation of the results is done through comparison of the segmentation quality with another similar alternative approach and it is found that the proposed approach is capable of producing more faithful segmentation.