Search

The cardiac magnetic resonance (CMR) images from a group of patients with myocardial scars and implanted cardioverter-defibrillator (ICD) are divided into a group with low risk of arrhythmias (late incidents) and a group with high risk of arrhythmias (early incidents). Several hundred quantitative features describing sizes, statistics and textures of the segmented and defined areas of the images are computed from manually segmented images in an exploratory analysis. The method used to determine decision regions to discriminate the patients with low risk of arrhythmias from the patient with high risk of arrhythmias is a maximum likelihood estimation based Bayes classifiers described in [1]. The results presented can be interpreted as hypothesis of which features, and combinations of features, that might have discriminative power. A major hypothesis that arises is that there are important textural information in the scarred and non-scarred areas.

Bone mineral density (BMD) is used in clinical medicine as an indirect indicator of osteoporosis and fracture risk. From a technical point of view Quantitative Computed Tomography (QCT) should be the gold standard in bone densitometry. On the other hand, it is known that a greater percentage increase in skin dose is needed as the patient size is increased: positive results and side effect of long-term steroid treatment as obesity have been found for Duchenne muscular dystrophy (DMD), characterized by a progressive muscle degeneration and substitution with fat. The present work is an effort to improve osteoporosis diagnostic efficacy in children by analyzing the trabecular bone texture in CT L3 vertebra by two methods which are independent of image intensity: fractal dimension with power spectrum and wavelet packets. As results, comparing healthy children (44 children both sexes) with osteoporotic subjects (13 adult women, aged 52–87 years) great differences were noticed in all image texture indicators (p

Snoring is often associated with serious health risks such as obstructive sleep apnea and heart disease and may require targeted surgical interventions. In this regard, research into automatically and unobtrusively analysing the site of blockages that cause snore sounds is growing in popularity. Herein, we investigate the use of low level image texture features in classification of four specific types of snore sounds. Specifically, we explore histogram of local binary patterns (LBP) in dense grid of rectangular regions and histogram of oriented gradients (HOG) extracted from colour spectrograms for snore sound characterisation. Support vector machines with homogeneous mapping are used in the classification stage of the proposed method. Various experimental works are carried out with both LBP and HOG descriptors on the INTERSPEECH ComParE 2017 snoring sub-challenge dataset. Results presented indicate that LBP descriptors are better than the HOG descriptors in snore type detection and fusion of the LBP and HOG descriptors produces stronger results than either individual descriptor. Further, when compared to the challenge baseline and state-of-the-art deep spectrum features, our approach achieved relative percentage increases in unweighted average recall of 23.1% and 8.3% respectively.

Non-invasive ultrasound imaging of carotid plaques can provide information on the characteristics of the arterial wall including the size, morphology and texture of the atherosclerotic plaques. Several studies were carried out that demonstrated the usefulness of these feature sets for differentiating between asymptomatic and symptomatic plaques and their corresponding cerebrovascular risk stratification. The aim of this study was to develop predictive modelling for estimating the time period of a stroke event by determining the risk for short term (less or equal to three years) or long term (more than three years) events. Data from 108 patients that had a stroke event have been used. The information collected included clinical and ultrasound imaging data. The prediction was performed at base line where patients were still asymptomatic. Several image texture analysis and clinical features were used in order to create a classification model. The different features were statistically analyzed and we conclude that image texture analysis features extracted using Spatial Gray Level Dependencies method had the best statistical significance. Several predictive models were derived based on Binary Logistic Regression (BLR) and Support Vector Machines (SVM) modelling. The best results were obtained with the SVM modelling models with an average correct classifications score of 77±7% for differentiating between stroke event occurrences within 3 years versus more than 3 years. Further work is needed in investigating additional multiscale texture analysis features as well as more modelling techniques on more subjects. © 2015 IEEE. 2015-November 334 337 Sponsors: Conference code: 116805

Image texture features are effective at characterizing the microstructure of cancerous tissues. This paper proposes predicting the survival times of glioblastoma multiforme (GBM) patients using texture features extracted in multi-contrast brain MRI images. Texture features are derived locally from contrast enhancement, necrosis and edema regions in T1-weighted post-contrast and fluid-attenuated inversion-recovery (FLAIR) MRIs, based on the gray-level co-occurrence matrix representation. A statistical analysis based on the Kaplan-Meier method and log-rank test is used to identify the texture features related with the overall survival of GBM patients. Results are presented on a dataset of 39 GBM patients. For FLAIR images, four features (Energy, Correlation, Variance and Inverse of Variance) from contrast enhancement regions and a feature (Homogeneity) from edema regions were shown to be associated with survival times (p-value < 0.01). Likewise, in T1-weighted images, three features (Energy, Correlation, and Variance) from contrast enhancement regions were found to be useful for predicting the overall survival of GBM patients. These preliminary results show the advantages of texture analysis in predicting the prognosis of GBM patients from multi-contrast brain MRI.

Contour tree represents the nesting relations of the individual components in the image; however, it neglects the geometric structure of the terrain. In this paper, we propose a new topological representation that provides the nesting and spatial relations of regions for CT image interpretation. The tree is constructed based on the signed distance transformation of binary CT image, and combines intensity based contour tree and a new gradient based topology tree. We also investigate the application of the topological representation as a constraint in target object segmentation. Ten non-small cell lung tumor CT studies were used for segmentation accuracy evaluation. The image representation results showed that the proposed tree structure retained the nesting and spatial relations of the tissues or objects in the CT image. The segmentation results demonstrated its usability in the separation of adjacent objects with similar intensity distributions.

Small structure segmentation from medical images is a challenging problem yet has important applications. Examples are labeling cell, lesion and glomeruli for disease diagnosis, just to name a few. Though extensive research has proposed various detectors for this type of problem, most are 2D detectors. Recently, we have developed a Hessian based 3D detector to segment small structures from medical images (e.g., MRI). In our detector, two 3D geometrical features: regional blobness and flatness, in conjunction with the intensity features are fully utilized to serve the segmentation purpose. The objective of this research is to further improve the 3D detector with additions of texture features. Medical images contain rich information which can be presented as texture, the local characteristics pattern of image intensity. We hypothesize the Hessian based detector extended with the 3D texture features is expected to have improved performance in segmenting small structures. To thoroughly evaluate the contributions from the textual features, 25 synthetic images and 6 real world rat MR images are studied. It is observed the combination of intensity, blobness, and two texture features: intensity standard deviation and entropy improves performance in synthetic dataset by about 19% in F-score, and performs as well as other detectors on rat MR images.

This paper proposes a novel texture sensing method for nursing-care gel by using an artificial mastication system, in which not only mechanical characteristics but also geometrical ones are objectively and quantitatively evaluated. When human masticates gel food, she or he perceives the changes of the shape and contact force simultaneously. Based on the impressions, they evaluate the texture. For reproducing such a procedure, the pressure distribution of gel is measured in the simple artificial mastication, and the information associated to both the geometrical and mechanical characteristics is simultaneously acquired. The relationship between the value of sensory evaluation (i.e. impression human perceives), and the pressure distribution data is numerically modeled by applying the image texture analysis. Experimental results show that the proposed method succeeds in estimating the values of sensory evaluation of nine kinds of gel with the coefficient of determination greater than 0.93.

Radiation dose of X-ray Computed Tomography (CT) imaging has raised a worldwide health concern. Therefore, low-dose CT imaging has been of a huge interest in the last decade. However, lowering the radiation dose degrades the image quality by increasing the noise level, which may reduce the diagnostic performance of the images. As a result, image denoising is one of the fundamental tasks in low-dose CT imaging. One of the state of art denoising methods, which has been successfully used in this area, is Total Variation (TV) denoising. Nevertheless, if the parameters of the TV denoising are not optimally adjusted or the algorithm is not stopped in an appropriate point, some of the small structures will be removed by this method. Here, we provide a solution to this problem by proposing a modified nonlocal TV method, called probabilistic NLTV (PNLTV). Denoising performance of PNLTV is improved by using better weights and an appropriate stopping criterion based on statistics of image wavelet coefficients. Non-locality allows the algorithm to preserve the image texture, which combined with the proposed stopping criterion enables PNLTV to keep fine details unchanged.

In the clinical environment, accuracy and speed of the image segmentation process plays a key role in the analysis of pathological regions. Despite advances in anatomic image segmentation, time-effective correction tools are commonly needed to improve segmentation results. Therefore, these tools must provide faster corrections with a low number of interactions, and a user-independent solution. In this work we present a new interactive correction method for correcting the image segmentation. Given an initial segmentation and the original image, our tool provides a 2D/3D environment, that enables 3D shape correction through simple 2D interactions. Our scheme is based on direct manipulation of free form deformation adapted to a 2D environment. This approach enables an intuitive and natural correction of 3D segmentation results. The developed method has been implemented into a software tool and has been evaluated for the task of lumbar muscle segmentation from Magnetic Resonance Images. Experimental results show that full segmentation correction could be performed within an average correction time of 6±4 minutes and an average of 68±37 number of interactions, while maintaining the quality of the final segmentation result within an average Dice coefficient of 0.92±0.03.

In this work, we have developed an adjunct Computer Aided Diagnostic (CAD) technique that uses 3D acquired ultrasound images of the ovary and data mining algorithms to accurately characterize and classify benign and malignant ovarian tumors. In this technique, we extracted image-texture based and Higher Order Spectra (HOS) based features from the images. The significant features were then selected and used to train and test the Decision Tree (DT) classifier. The proposed technique was validated using 1000 benign and 1000 malignant images, obtained from 10 patients with benign and 10 with malignant disease, respectively. On evaluating the classifier with 10-fold stratified cross validation, we observed that the DT classifier presented a high accuracy of 95.1%, sensitivity of 92.5% and specificity of 97.7%. Thus, the four significant features could adequately quantify the subtle changes and nonlinearities in the pixel intensities. The preliminary results presented in this paper indicate that the proposed technique can be reliably used as an adjunct tool for ovarian tumor classification since the system is accurate, completely automated, cost-effective, and can be easily written as a software application for use in any computer.

Wireless endoscopy systems is a new innovative method widely used for gastrointestinal tract examination in recent decade. Wireless Micro-Ball endoscopy system with multiple image sensors is the newest proposed method which can make a full view image of the gastrointestinal tract. But still the quality of images from this new wireless endoscopy system is not satisfactory. It's hard for doctors and specialist to easily examine and interpret the captured images. The image features also are not distinct enough to be used for further processing. So as to enhance these low-contrast endoscopic images a new image enhancement method based on the endoscopic images features and color distribution is proposed in this work. The enhancement method is performed on three main steps namely color space transformation, edge preserving mask formation, and histogram information correction. The luminance component of CIE Lab, YCbCr, and HSV color space is enhanced in this method and then two other components added finally to form an enhanced color image. The experimental result clearly show the robustness of the method. © 2014 IEEE.

Multi-modal image registration has been a challenging task in medical images because of the complex intensity relationship between images to be aligned. Registration methods often rely on the statistical intensity relationship between the images which suffers from problems such as statistical insufficiency. The proposed registration method works based on extracting structural features by utilizing the complex phase and gradient-based information. By employing structural relationships between different modalities instead of complex similarity measures, the multi-modal registration problem is converted into a mono-modal one. Therefore, conventional mono-modal similarity measures can be utilized to evaluate the registration results. This new registration paradigm has been tested on magnetic resonance (MR) brain images of different modes. The method has been evaluated based on target registration error (TRE) to determine alignment accuracy. Quantitative results demonstrate that the proposed method is capable of achieving comparable registration accuracy compared to the conventional mutual information.

In this work, an attempt has been made to perform asymmetry analysis in breast thermograms using non-linear total variation diffusion filter and reaction diffusion based level set method. Breast images used in this study are obtained from online database of the project PROENG. Initially the images are subjected to total variation (TV) diffusion filter to generate the edge map. Reaction diffusion based level set method is employed to segment the breast tissues using TV edge map as stopping boundary function. Asymmetry analysis is performed on the segmented breast tissues using wavelet based structural texture features. The results show that nonlinear total variation based reaction diffusion level set method could efficiently segment the breast tissues. This method yields high correlation between the segmented output and the ground truth than the conventional level set. Structural texture features extracted from the wavelet coefficients are found to be significant in demarcating normal and abnormal tissues. Hence, it appears that the asymmetry analysis on segmented breast tissues extracted using total variation edge map can be used efficiently to identify the pathological conditions of breast thermograms.

To support 3D magnetic resonance image (MRI) analysis, a marginal image similarity (MIS) matrix holding MR inter-slice relationship along every axis view (Axial, Coronal, and Sagittal) can be estimated. However, mutual inference from MIS view information poses a difficult task since relationships between axes are nonlinear. To overcome this issue, we introduce a Tensor-Product Kernel-based Representation (TKR) that allows encoding brain structure patterns due to patient differences, gathering all MIS matrices into a single joint image similarity framework. The TKR training strategy is carried out into a low dimensional projected space to get less influence of voxel-derived noise. Obtained results for classifying the considered patient categories (gender and age) on real MRI database shows that the proposed TKR training approach outperforms the conventional voxel-wise sum of squared differences. The proposed approach may be useful to support MRI clustering and similarity inference tasks, which are required on template-based image segmentation and atlas construction.

Kidney segmentation is an important step for computer-aided diagnosis or treatment in urology. In this paper, we present an automatic method based on multi-atlas image registration for kidney segmentation. The method mainly relies on a two-step framework to obtain coarse-to-fine segmentation results. In the first step, down-sampled patient image is registered with a set of low-resolution atlas images. A coarse kidney segmentation result is generated to locate the left and right kidneys. In the second step, the left and right kidneys are cropped from original images and aligned with another set of high-resolution atlas images to obtain the final results respectively. Segmentation results from 14 CT angiographic (CTA) images show that our proposed method can segment the kidneys with a high accuracy. The average Dice similarity coefficient and surface-to-surface distance between segmentation results and reference standard are 0.952 and 0.913mm. Furthermore, the kidney segmentation in CT urography (CTU) and CTA images of 12 patients were performed to show the feasibility of our method in CTU images.

The application of multimodal image registration to various medical applications has been investigated. Image fusion involving 3D thermal and MRI/CT images allows the extraction of both functional and anatomical information, which may become a powerful tool to aid in clinical diagnoses. This paper presents innovations at the image fusion methodology, which currently requires that both imaging modalities are represented and visualized at the same 3D viewing projection. The proposed solution is based and compared with two different viewing projections: orthogonal and perspective. The methodology requires that the thermographic images (or photographs) are visualized in the orthogonal view, in order to match with the 2D projected images (using range images) from MRI/CT. The results obtained have shown significant improvements in the 3D thermal models, when compared and evaluated with the perspective approach. This allowed the generation of more accurate 3D models, which match both the geometry and texture (functional temperature information). Since it is desirable to combine or unify more than one imaging modality, these 3D multimodal models may have a strong impact in many clinical applications.

Manual segmentation of ultrasound contrast images is time-consuming and inevitable to variability, and computer-based segmentation algorithms often require user interaction. This paper proposes a novel level set model for fully automated segmentation of vascular ultrasound contrast images. The initial contour of arterial boundaries is acquired based on an automatic procedure. The level set model moves the initial contour towards the boundaries of arterial inner wall based on minimization of the energy function. The traditional energy function is improved by introducing an edge detector based on image gradient and the standard difference image. Both spatial and temporal information of the image are considered, and the robustness and accuracy of the level set model is enhanced. Ultrasonic contrast images of living mouse are acquitted with high frequency ultrasound system. Images of carotid arteries are processed with our method. The segmentation results using the proposed method are evaluated against two observers' hand-outlined boundaries, showing that computer-generated boundaries agree well with the observers' hand-outlined boundaries as much as the different observers agree with each other.

Ultrasound lesion segmentation is an important and challenging task. Comparing with other methods, region-based level set has many advantages, but still requires considerable improvement to deal with the characteristic of lesions in the ultrasound modality such as shadowing, speckle and heterogeneity. In the clinical workflow, the physician would usually denote long and short axes of a lesion for measurement purpose yielding four markers in an image. Inspired by this workflow, a constrained level set method is proposed to fully utilize these four markers as prior knowledge and global constraint for the segmentation. First, the markers are detected using template-matching algorithm and B-Spline is applied to fit four markers as the initial contour. Then four-marker constrained energy is added to the region-based local level set to make sure that the contour evolves without deviation from the four markers. Finally the algorithm is implemented in a multi-resolution scheme to achieve sufficient computational efficiency. The performance of the proposed segmentation algorithm was evaluated by comparing our results with manually segmented boundaries on 308 ultrasound images with breast lesions. The proposed method achieves Dice similarity coefficient 89.49 ± 4.76% and could be run in real-time.

We propose a novel multiresolution framework for ultrasound image segmentation in this paper. The framework exploits both local intensity and local phase information to tackle the degradations of ultrasound images. First, multiresolution scheme is adopted to build a Gaussian pyramid for each speckled image. Speckle noise is gradually smoothed out at higher levels of the pyramid. Then local intensity-driven active contours are employed to locate the coarse contour of the target from the coarsest image, followed by local phase-based geodesic active contours to further refine the contour in finer images. Compared with traditional gradient-based methods, phase-based methods are more suitable for ultrasound images because they are invariant to variations in image contrast. Experimental results on left ventricle segmentation from echocardiographic images demonstrate the advantages of the proposed model.

This paper focuses on breast self-examination (BSE) stroke position and palpation level classification for the development of a computer vision-based BSE training and guidance system. In this study, image frames are extracted from a BSE video and processed considering the color information, shape, and texture by wavelet transform and first order color moment. The new approach using artificial neural network and wavelet transform can identify BSE stroke positions and palpation levels, i.e. light, medium, and deep, at 97.8 % and 87.5 % accuracy respectively.

The performance of automated analysis of cellular images is heavily influenced by the features that characterize cells or cell nuclei. In this paper, an exhaustive set of features including morphological, topological, and texture features are explored to determine the optimal features for classification of cells and cell nuclei. The optimal subset of features are obtained using popular feature selection methods. The results of feature selection indicate that Zernike moment, Daubechies wavelets, and Gabor wavelets give the most important features for the classification of cells or cell nuclei in fluorescent microscopy images.

Presence of peripapillary chorioretinal atrophy (PPA) is considered one of the risk factors for glaucoma. It can be identified as bright regions in retinal fundus images, and therefore, incorrectly included as the part of the optic disc regions in the automated disc detection scheme. For potential risk assessment and use in improving optic disc segmentation, a computerized detection of PPA was investigated. By using texture analysis, the sensitivity for detecting the moderate to severe PPA regions in the test dataset was 73% with the specificity of 95%. The proposed method may be useful for identifying the cases with the PPA in retinal fundus images.

Cataract remains a leading cause for blindness worldwide. Cataract diagnosis via human grading is subjective and time-consuming. Several methods of automatic grading are currently available, but each of them suffers from some drawbacks. In this paper, a new approach for automatic detection based on texture and intensity analysis is proposed to address the problems of existing methods and improve the performance from three aspects, namely ROI detection, lens mask generation and opacity detection. In the detection method, image clipping and texture analysis are applied to overcome the over-detection problem for clear lens images and global thresholding is exploited to solve the under-detection problem for severe cataract images. The proposed method is tested on 725 retro-illumination lens images randomly selected from a database of a community study. Experiments show improved performance compared with the state-of-the-art method.

Breast density has been shown to be one of the most significant risks for developing breast cancer, with women with dense breasts at four to six times higher risk. The Breast Imaging Reporting and Data System (BI-RADS) has a four class classification scheme that describes the different breast densities. However, there is great inter and intra observer variability among clinicians in reporting a mammogram's density class. This work presents a novel texture classification method and its application for the development of a completely automated breast density classification system. The new method represents the mammogram using textons, which can be thought of as the building blocks of texture under the operational definition of Leung and Malik as clustered filter responses. The new proposed method characterizes the mammographic appearance of the different density patterns by evaluating the texton spatial dependence matrix (TDSM) in the breast region's corresponding texton map. The TSDM is a texture model that captures both statistical and structural texture characteristics. The normalized TSDM matrices are evaluated for mammo-grams from the different density classes and corresponding texture models are established. Classification is achieved using a chi-square distance measure. The fully automated TSDM breast density classification method is quantitatively evaluated on mammograms from all density classes from the Oxford Mammogram Database. The incorporation of texton spatial dependencies allows for classification accuracy reaching over 82%. The breast density classification accuracy is better using texton TSDM compared to simple texton histograms. &nbsp

Computer Tomographic Colonography, combined with computer-aided detection (CAD), is a promising emerging technique for colonic polyp analysis. We present a CAD scheme for polyp flagging based on new texture and geometric features that consider both the information in the candidate polyp location and its immediate surrounding area, testing multiple sizes. The proposed algorithm is tested with ground truth data, including flat and small polyps, with very promising results.

This paper is concerned with the classification of tumoral tissue in the small bowel by using capsule endoscopic images. The followed approach is based on texture classification. Texture descriptors are derived from selected scales of the Discrete Curvelet Transform (DCT). The goal is to take advantage of the high directional sensitivity of the DCT (16 directions) when compared with the Discrete Wavelet Transform (DWT) (3 directions). Second order statistics are then computed in the HSV color space and named Color Curvelet Covariance (3C) coefficients. Finally, these coefficients are modeled by a Gaussian Mixture Model (GMM). Sensitivity of 99% and specificity of 95.19% are obtained in the testing set.

We present a technique for automatic diagnosis of malignant melanoma based exclusively on local pattern analysis. The technique relies on local binary patterns in small sections in the image, and automatically selects the relevant texture features from those that discriminate best between benign and malignant skin lesions. The classification is performed using support vector machines, and the feature vectors are clustered using K-means clustering. The effects of K and window size are investigated. Reported average specificity and sensitivity are 73% for optimal parameter choice, indicating that the procedure is a useful part of a diagnostic system.

In this paper, we present a method that detects intracranial space-occupying lesions in two-dimensional (2D) brain high-resolution CT images. Use of statistical texture atlas technique localizes anatomy variation in the gray level distribution of brain images, and in turn, identifies the regions with lesions. The statistical texture atlas involves 147 HRCT slices of normal individuals and its construction is extremely time-consuming. To improve the performance of atlas construction, we have implemented the pixel-wise texture extraction procedure on Nvidia 8800GTX GPU with Compute Unified Device Architecture (CUDA) platform. Experimental results indicate that the extracted texture feature is distinctive and robust enough, and is suitable for detecting uniform and mixed density space-occupying lesions. In addition, a significant speedup against straight forward CPU version was achieved with CUDA.

In this paper, a comparison is made between a global and a local approach for computer-aided classification of virtual slides from breast tumor sections. The first approach classifies the images according to both color and texture information from the whole tissue, whereas the second one classifies them only on shape and texture from epithelial compartment. The originality of this study is that only low resolution virtual slides are used. As the cytological details are not visible at low resolution, the classification is only based on architectural aspect of each lesion. Experimental results on images from breast tumor sections show that some information is already present in low resolution virtual slides, allowing the classification in benign lesions versus malignant carcinomas.

Flexible endoscopes based on fiber bundles are still widely used despite the recent success of so-called tipchip endoscopes. This is partly due to the costs and that for extremely thin diameters (below 3 mm) there are still only fiberscopes available. Due to the inevitable artifacts caused by the transition from the fiber bundles to the sensor chip, image and texture analysis algorithms are severely handicapped. Therefore, texture-based computer-assisted diagnosis (CAD) systems could not be used in such domains without image preprocessing. We describe a CAD system approach that includes an image filtering algorithm to remove the fiber image artifacts first and then applies conventional color texture algorithms that have been applied to other endoscopic disciplines in the past. The concept is evaluated on an image database with artificially rendered fiber artifacts so that ground truth information is available.

Navigation during Minimally Invasive Surgery (MIS) has recognized difficulties due to limited field-of-view, off-axis visualization and loss of direct 3D vision. This can cause visual-spatial disorientation when exploring complex in vivo structures. In this paper, we present an approach to dynamic view expansion which builds a 3D textured model of the MIS environment to facilitate in vivo navigation. With the proposed technique, no prior knowledge of the environment is required and the model is built sequentially while the laparoscope is moved. The method is validated on simulated data with known ground truth. Its potential clinical value is also demonstrated with in vivo experiments.

Tissue classification in mammography can help the diagnosis of breast cancer by separating healthy tissue from lesions. We present herein the use of three texture descriptors for breast tissue segmentation purposes: the Sum Histogram, the Gray Level Co-Occurrence Matrix (GLCM) and the Local Binary Pattern (LBP). A modification of the LBP is also proposed for a better distinction of the tissues. In order to segment the image into its tissues, these descriptors are compared using a fidelity index and two clustering algorithms: k-Means and SOM (Self-Organizing Maps).

Traditional endoscopic methods do not allow the visualization of the entire Gastrointestinal (GI) tract. Wireless Capsule Endoscopy (CE) is a diagnostic procedure that overcomes this limitation of the traditional endoscopic methods. The CE video frames possess rich information about the condition of the stomach and intestine mucosa, encoded as color and texture patterns. It is known for a long time that human perception of texture is based in a multi-scale analysis of patterns, which can be modeled by multi-resolution approaches. Furthermore, modeling the covariance of textural descriptors has been successfully used in classification of colonoscopy videos. Therefore, in the present paper it is proposed a frame classification scheme based on statistical textural descriptors taken from the Discrete Curvelet Transform (DCT) domain, a recent multi-resolution mathematical tool. The DCT is based on an anisotropic notion of scale and high directional sensitivity in multiple directions, being therefore suited to characterization of complex patterns as texture. The covariance of texture descriptors taken at a given detail level, in different angles, is used as classification feature, in a scheme designated as Color Curvelet Covariance. The classification step is performed by a multilayer perceptron neural network. The proposed method has been applied in real data taken from several capsule endoscopic exams and reaches 97.2% of sensitivity and 97.4% specificity. These promising results support the feasibility of the proposed method.

This paper presents color image processing methods for the analysis of dermatological images in the context of a content-based image retrieval (CBIR) system. Tests were conducted on the classification of tissue components in skin lesions, in terms of necrotic tissue, fibrin, granulation, and mixed composition. The images were classified based on color components by an expert dermatologist following a black-yellow-red model. Indexing and retrieval of images were performed based on texture information obtained from the red, green, blue, hue, and saturation components of the color images. The performance of the CBIR system was measured in terms of precision and recall. Initial results demonstrate the potential of the proposed methods with the best precision result of 70% obtained for the characterization of mixed tissue composition.

Carefully and appropriately designed texture patterns, for highlighting the underlying geometry, in addition to standard shading and simulated lighting, can enhance the perception of 3D surface properties such as curvature and orientation. In this paper we propose enhancements to a texture synthesis algorithm for 3D medical data visualization, incorporating a robust texture patch boundary handling technique. We compare different texture and color mapping techniques on a medical object and evaluate their effects based on feedback from human observers. Experimental results show the feasibility of our enhanced algorithm.

The use of the watershed algorithm for image segmentation is widespread because it is able to produce a complete division of the image. However, it is susceptible to over-segmentation and in medical image segmentation, this meant that that we do not have good representations of the anatomy. We address this issue by thresholding the gradient magnitude image and performing post-segmentation merging on the initial segmentation map. The automated thresholding technique is based on the histogram of the gradient magnitude map while the post-segmentation merging is based on the similarity in textural features (namely angular second moment, contrast, entropy and inverse difference moment) belonging to two neighboring partitions. When applied to the segmentation of various facial anatomical structures from magnetic resonance (MR) images, the proposed method achieved an overlap index of 92.6% compared to manual contour tracings. It is able to merge more than 80% of the initial partitions, which indicates that a large amount of over-segmentation has been reduced. Results produced using watershed algorithm with and without the proposed and proposed post-segmentation merging are presented for comparisons.

The objective of this study was to develop a CAD system for the classification of hysteroscopy images of the endometrium based on color texture analysis for the early detection of gynaecological cancer. A total of 416 Regions of Interest (ROIs) of the endometrium were extracted (208 normal and 208 abnormal) from 40 subjects. RGB images were gamma corrected and were converted to the HSV and YCrCb color systems. The following texture features were extracted for each channel of the RGB, HSV, and YCrCb systems: (1) Statistical Features, (ii) Spatial Gray Level Dependence Matrices and (iii) Gray Level Difference Statistics. The PNN statistical learning and SVM neural network classifiers were also investigated for classifying normal and abnormal ROIs. Results show that there is significant difference (using the Wilcoxon Rank Sum Test at a=0.05) between the texture features of normal and abnormal ROIs of the endometrium. Abnormal ROIs had higher gray scale median, variance, entropy and contrast and lower gray scale median and homogeneity values when compared to the normal ROIs. The highest percentage of correct classifications score was 79% and was achieved for the SVM models trained with the SF and GLDS features for differentiating between normal and abnormal ROIs. Concluding, a CAD system based on texture analysis and SVM models can be used to classify normal and abnormal endometrium tissue. Further work is needed to validate the system in more cases and organs. © 2007 IEEE. 864 867 Sponsors: ACIES, Research Promotion and Management Consulting EOARD, European Office of Aerospace R and D Grand Lyon NSF, National Science Foundation Philips Research Europe Philips Research North America Conference code: 70818 Cited By :6

Analysis of ultrasound speckle texture will provide us information about the underlying properties of tissue, could find applications in early lesion detection and tissue characterization. Traditional first and second order statistics based approaches ignore the higher order statistics information in the texture. On the other hand, conventional multichannel filtering or multiresolution analysis approaches rely on the predefined analytical bases which are not fully adaptive to the data being analyzed. In this paper independent component analysis (ICA), which is based on higher order statistics, is proposed to deal with the ultrasound speckle texture analysis problem. ICA image bases obtained from the training images are applied as a filter bank to the testing images. Then the independent features containing higher order statistics information can be extracted from the marginal distributions of the filtered images. ICA is used here as a dimensionality reduction tool to overcome the difficulty of estimating high dimensional joint density of texture. Support Vector Machine (SVM) is then used as a classifier to classify the tissues. By using the digitally simulated tissues and corresponding B-scan images, we can further correlate the change of tissue microstructure or change of imaging conditions with the change of the ICA feature vectors. Our numerical simulation has shown ICA to be a promising technique for ultrasound speckle texture analysis and tissue characterization compared with some traditional methods such as PCA and Gabor transform.

The aging population in developed countries has shifted considerable research attention to diseases related to age. Because age is one of the highest risk factors for neurodegenerative diseases, the need for automated brain image analysis has significantly increased. Magnetic resonance imaging (MRI) is a commonly used modality to image brain. MRI provides high tissue contrast; hence, the existing brain image analysis methods have often preferred the intensity information to others, such as texture. Recently, an easy-to- compute texture descriptor, local binary pattern (LBP), has shown promise in various applications outside the medical field. In this paper, after extensive experiments, we show that rotation-invariant LBP is invariant to some common MRI artifacts that makes it possible to use it in various high-level brain MR image analysis applications.

An accurate segmentation of solid nodules in ultrasonographic (US) breast image is presented. 1-level 2-dimensional Discrete Wavelet Transform (DWT) is used to create features reflecting the texture information of the original image. Using these features, the texture classification is achieved. Finally, solid nodule region is segmented from the classified texture region. Proper threshold for texture classification is automatically decided. Empirically acquired information about the relationship between the texture characteristic of the original image and the optimal threshold is examined and used. Presented algorithm is applied to 284 malignant solid nodules and 300 benign solid nodules and the resulting images are presented.

Prostate cancer diagnosis and treatment rely on segmentation of Transrectal Ultrasound (TRUS) prostate images. This is a challenging and difficult task dut to weak prostate boundaries, speckle noise and the short range of gray levels. Advances in digital imaging techniques have made it possible the acquisition of large volumes of TRUS prostate images so that there is considerable demand for automated segmentation systems. Local binary pattern (LBP) has been used for texture segmentation and analysis. Despite its promising performance for texture classification it has not yet been considered for TRUS prostate segmentation. In this paper we introduce a medical texture local binary pattern operator designed for applications of medical imaging where different tissues or micro organisms might maintain extremely weak underlying textures that make it impossible or very difficult for ordinary texture analysis approaches to classify them. In the proposed method the deformations of a level set contour are controlled based on the medical texture local binary pattern operator.

Bone architectural information can be derived from structural indices calculated from high resolution magnetic resonance (MR) images. However, high resolution scans are time consuming and prone to motion artifacts and hence are not routinely feasible. The purpose of this study is to determine if a correlation exists between 3D structural indices calculated from high-resolution MR images and 3D co-occurrence texture features calculated from lower resolution MR images. Regression analysis indicates a strong correlation between the structural indices and the texture features. This study highlights the potential of using surrogate texture markers extracted from readily acquired clinical MR images to quantify bone architecture, circumventing the need for high resolution MR imaging.

Ultrasound Medical Imaging is currently the most popular modality for diagnostic application. This imaging technique has been used for the detecting abnormalities associated with abdominal organs like liver, kidney, uterus etc. In this paper, the possibilities of automatic classification of the ultrasound liver images into four classes-normal, cyst, benign and malignant masses, using texture features are explored. These texture features are extracted using the various statistical and spectral methods. The optimal feature selection process is carried out manually to pick the best discriminating features from the extracted texture parameters. Also, the method of principal component analysis is used to extract the principal features or directions of maximum information from the data set there by automatically selecting the optimal features. Using these optimal features, a final combined feature set is formed and is employed for classification of the liver lesions into respective classes. K-means clustering and neural network based automatic classifiers are employed in this process. The classifier design and its performance are studied. This paper summarizes the various statistical and spectral texture parameter extraction processes, optimal feature selection techniques and automated classification procedures involved in our work.