Statistical Texture Descriptors Research Articles

Characterization and Classification of Brain Tissue and Stroke Lesions in Non-Contrast Computed Tomography Images of Stroke Patients Using Statistical Texture Descriptors and Artificial Neural Network

 
 
 
 
 
 Aim: To characterize and classify stroke lesions and normal brain tissue in computed tomography (CT) images using statistical texture descriptors.
 Patients and methods: Two experienced radiologists blinded to each other inspected CT images of 164 stroke patients to identify and categorize stroke lesions into ischaemic and haemorrhagic subtypes. Four regions of interest (ROIs) in each CT slice that demonstrated the lesion; two each representing the lesion and normal tissue were selected. Statistical texture descriptors namely, co-occurrence matrix, run-length matrix, absolute gradient and histogram were calculated for them.  Raw data analysis was performed to identify the parameters that best discriminate between normal brain tissue and stroke lesions. Artificial neural network (ANN) was used to classify the ROIs into normal tissue, ischaemic and haemorrhagic lesions using the radiologists’ identification and categorization as the gold standard, and further analyzed using the receiver operating characteristic curve.
 Results: Three parameters in each texture class discriminated between normal tissue, ischaemic and haemorrhagic stroke lesions. The discriminating co-occurrence matrix parameters were sum average parameters namely S1-1 SumAverg, S1-0 SumAverg and S0-1 SumAverg.  For the run-length matrix, short run emphasis in horizontal, 1350 and 450 directions were the discriminating features. The discriminating absolute gradient parameters were gradient non-zeros, gradient variance and gradient mean. For the histogram class, the mean, 90th and 99th percentiles were the discriminating parameters. The ANN achieved a sensitivity of 0.637, specificity 0.753, false positive rate (FPR) 0.247, and false negative rate (FNR) 0.363 with co-occurrence matrix. With run-length matrix the sensitivity was 0.544, specificity 0.607, FPR 0.393, and FNR 0.456 while with absolute gradient the sensitivity was 0.546, specificity 0.586, FPR 0.414, FNR 0.454. With histogram, the sensitivity was 0.947, specificity 0.962, FPR 0.038, and FNR 0.053.
 Conclusion: The histogram texture features showed the highest sensitivity and specificity in the classification of brain tissue and stroke lesions using the artificial neural network. 
  
 
 
 
 
  

New local binary pattern approaches based on color channels in texture classification

In this paper, four novel, simple and robust approaches, which are left to right local binary patterns (LBPLL2R), top to down local binary patterns (LBPT2D), cube surface local binary pattern (LBPSurfaces), and cube diagonal local binary pattern (LBPDiagonal), were proposed in order to exact texture features in color images. These approaches were based on the local binary pattern (LBP), which is an effective statistical texture descriptor and can be employed in gray images. Proposed approaches were evaluated and validated in four datasets, which are Outex, KTH_TIPS, KTH_TIPS2, and USPtex datasets. The images in these datasets are in RGB, HSV, YIQ, and YCbCr color formats. Achieved results by these approaches were compared with the obtained results by the classical LBP and literature findings. As a result, the proposed approaches performed better than the traditional LBP method and they found effective in the classification of color texture images, especially in images, which are in RGB and HSV formats. Furthermore, noise robustness and time complexity of the proposed approaches were validated.

Computer Aided Diagnosis of Abdomen Diseases Using Curvelet Transform

Abdomen related diseases are responsible of many deaths every year. These deaths can be reduced by early diagnosis of abdomen diseases. Computer aided diagnosis (CAD) can play vital role in early detection of diseases. Hence, a novel CAD is proposed in this paper that can diagnose abdomen diseases like Hepatocellular carcinoma, cysts and Calculi using statistical curvelet texture descriptors. The proposed CAD is divided into four stages: (a) Image segmentation using active contours, (b) feature extraction, (c) feature selection and (d) abdomen disease classification. The regions of interest (ROIs) are segmented from 120[Formula: see text]CT images using active contour models. The statistical features are extracted from segmented ROIs. Further, the classifiers are used to evaluate the ability of feature set in diagnosis various diseases of abdomen. The performance metrics indicates that the proposed CAD achieves accuracy of 87.9% using curvelet coefficient features and neural network as classifier.

Computer aided diagnosis system for abdomen diseases in computed tomography images

In this paper, a computer aided diagnostic (CAD) system for classification of abdomen diseases from computed tomography (CT) images is presented. The methodology used in this paper is to select the most appropriate machine learning technique of segmentation, feature extraction and classification for each module of proposed CAD. The methodology of selecting appropriate machine learning technique for each module of CAD results in accurate and efficient system. Regions of interest are segmented from CT images of tumor, cyst, calculi and normal liver using active contour models, region growing and thresholding. The CAD presented in this research work exploits the discriminating power of features for classifying abdominal diseases. Therefore, feature extraction module extracts statistical texture descriptors using three kinds of feature extraction methods i.e. Gray-Level co-occurrence matrices (GLCM), Discrete Wavelet Transform (DWT) and Discrete Curvelet Transform (DCT). At the next stage, effective and optimum features of ROIs are selected using Genetic Algorithm (GA). Further, Support Vector Machine (SVM) and Artificial Neural Network (ANN) are used to assess the capability of features for classification of diseases of abdomen. The study is performed on 120 CT images of abdomen (30 normal, 30 tumor, 30 cyst and 30 calculi). It is observed from the results that proposed CAD consists of edge based active contour model combined with optimized statistical texture descriptors using DCT along with ANN as classifier achieves the best diagnostic performance of 95.1%. It is also shown in results that proposed CAD achieves highest sensitivity, specificity of 95% and 98% respectively.

On The Effect Of Image Brightness And Contrast Nonuniformity On Statistical Texture Parameters

Abstract Computerized texture analysis characterizes spatial patterns of image intensity, which originate in the structure of tissues. However, a number of texture descriptors also depend on local average image intensity and/or contrast. This variations, known as image nonuniformity (inhomogeneity) artefacts often occur, e.g. in MRI. Their presence may lead to errors in tissue description. This unwanted effect is explained in this paper using statistical texture descriptors applied for MRI slices of a normal and fibrotic liver. To reduce the errors, correction of image spatial nonuniformity prior to texture analysis is performed. The issue of sensitivity of popular texture parameters to image nonuniformities is discussed. It is illustrated by classification examples of natural Brodatz textures, digitally modified to account for inhomogeneities – modeled as smooth variations of image intensity and contrast. A set of texture features is identified which represent certain immunity to image inhomogeneities.

Two novel local binary pattern descriptors for texture analysis

The recent developments in the image quality, storage and data transmission capabilities increase the importance of texture analysis, which plays an important role in computer vision and image processing. Local binary pattern (LBP) is an effective statistical texture descriptor, which has successful applications in texture classification. In this paper, two novel descriptors were proposed to search different patterns in images built on LBP. One of them is based on the relations between the sequential neighbors with a specified distance and the other one is based on determining the neighbors in the same orientation through central pixel parameter. These descriptors are tested with the Brodatz-1, Brodatz-2, Butterfly and Kylberg datasets to show the applicability of the proposed nLBPd and dLBPα descriptors. The proposed methods are also compared with classical LBP. The average accuracies obtained by ANN with 10 fold cross validation, which are 99.26% (LBPu2 and nLBPd), 94.44% (dLBPα), 95.71% (nLBPdu2) and %99.64 (nLBPd), for Brodatz-1, Brodatz-2, Butterfly and Kylberg datasets, respectively, show that the proposed methods outperform significant accuracies.

Performance Analysis of Discrete Wavelet Transform Based First-order Statistical Texture Features for Hardwood Species Classification

Abstract A simple and efficient discrete wavelet transform (DWT) based first-order statistical (FOS) texture descriptor is proposed in this paper to accurately classify the microscopic images of hardwood species. Primarily, DWT decomposes each image up to 8 levels using selected Daubechies (db1- db10) wavelet as a decomposition filter. Subsequently, four FOS features, namely, mean, standard deviation, skewness and kurtosis are employed to obtain substantial signatures of these images at different levels. The db3 based FOS texture features has achieved 96.80% classification accuracy compared to 93.20% classification accuracy obtained by local binary pattern features using linear support vector machine (SVM) classifier.

An Efficient Texture Classification Technique Based on Semi Uniform LBP

Local Binary Pattern' (LBP) operator has been proved to be a theoretically simple yet very effective statistical texture descriptor in many areas. One of the main disadvantage of LBP is they are more sensitive to even a simple noise, this causes a change in LBP pattern. A LBP can generate 256 (0 to 255) different patterns. These patterns are grouped into 59 uniform and 197 nonuniform LBP patterns. Most of the researchers have considered only uniform patterns for texture classification due to their small numbers and claiming that most of the textures are dominated by only uniform LBP's. But after a careful, critical and significant study on the literature, ULBP have some shortcomings: they discard some important texture information, suffer much from non-monotonic illumination variation and do not describe the stochastic characteristics of texture efficiently and sensitive to noise. To overcome this, the present paper defined 'Semi Uniform LBP' (SULBP). The proposed

An automatic computer-aided diagnosis system for liver tumours on computed tomography images

Liver cancer, one of the more common cancer diseases that cause a large number of deaths every year, can be reduced by early detection and diagnosis. Computer-Aided Diagnosis (CAD) can play a key role in the early detection and diagnosis of liver cancer. This paper develops a novel computer-aided diagnosis system focussing on the discriminating power of statistical texture descriptors in characterizing hepatocellular (malignant) from hemangioma (benign) liver tumours. The CAD system consists of three stages: (i) automatic tumour segmentation, (ii) texture feature extraction and (iii) tumour characterization using a classifier. Specifically, four features sets, the original gray level; co-occurrence of gray level; wavelet coefficient statistics and contourlet coefficient statistics are extracted from the tumour region of interest. A probabilistic neural network classifier is used to investigate the ability of each feature set in differentiating malignant from benign tissues. The performance of the CAD system evaluated using a database of images indicates that the highest accuracy achieved is 96.7% and the highest sensitivity and specificity are 97.3% and 96%, respectively that had been obtained with the contourlet coefficient co-occurrence features. The experimental results suggest that the developed CAD system has great potential and promise in the automatic diagnosis of both benign and malignant tumours of liver.

Accurate Prostate Volume Estimation Using Multifeature Active Shape Models on T2-weighted MRI

Accurate prostate volume estimation is useful for calculating prostate-specific antigen density and in evaluating posttreatment response. In the clinic, prostate volume estimation involves modeling the prostate as an ellipsoid or a spheroid from transrectal ultrasound, or T2-weighted magnetic resonance imaging (MRI). However, this requires some degree of manual intervention, and may not always yield accurate estimates. In this article, we present a multifeature active shape model (MFA) based segmentation scheme for estimating prostate volume from in vivo T2-weighted MRI. We aim to automatically determine the location of the prostate boundary on in vivo T2-weighted MRI, and subsequently determine the area of the prostate on each slice. The resulting planimetric areas are aggregated to yield the volume of the prostate for a given patient. Using a set of training images, the MFA learns the most discriminating statistical texture descriptors of the prostate boundary via a forward feature selection algorithm. After identification of the optimal image features, the MFA is deformed to accurately fit the prostate border. An expert radiologist segmented the prostate boundary on each slice and the planimetric aggregation of the enclosed areas yielded the ground truth prostate volume estimate. The volume estimation obtained via the MFA was then compared against volume estimations obtained via the ellipsoidal, Myschetzky, and prolated spheroids models. We evaluated our MFA volume estimation method on a total 45 T2-weighted in vivo MRI studies, corresponding to both 1.5 Tesla and 3.0 Tesla field strengths. The results revealed that the ellipsoidal, Myschetzky, and prolate spheroid models overestimated prostate volumes, with volume fractions of 1.14, 1.53, and 1.96, respectively. By comparison, the MFA yielded a mean volume fraction of 1.05, evaluated using a fivefold cross-validation scheme. A correlation with the ground truth volume estimations showed that the MFA had an r(2) value of 0.82, whereas the clinical volume estimation schemes had a maximum value of 0.70. Our MFA scheme involves minimal user intervention, is computationally efficient and results in volume estimations more accurate than state of the art clinical models.

A novel extended local-binary-pattern operator for texture analysis

The local-binary-pattern (LBP) operator has been proved to be a theoretically simple yet very effective multiresolution statistical texture descriptor in terms of the characteristics of the local structure, and has been applied in many areas. However, the extension to the LBP operator based on “uniform” patterns still has some shortcomings: it discards some important texture information and is sensitive to noise. A novel extended LBP operator for texture analysis is proposed in this paper. The new LBP operator classifies and combines the “nonuniform” local patterns based on analyzing their structure and occurrence probability. The new operator fully uses the texture information contained in the “nonuniform” local patterns, which is discarded by the classical LBP operators, and then becomes more robust against noise. Three experiments on the Brodatz texture database show the performance improvement of this novel extended LBP operator.

Increased potential for tissue discrimination in quantitative magnetic resonance imaging

A method has been developed that substantially increases quantitative tissue discrimination of pathology in in-vivo magnetic resonance (MR) imaging. Results from a study of carcinoma and benign hypertrophy of the prostate show the potential for classification of tissues from individual patients and demonstrate increased class separation. This appears to be the firstapplication of pattern recognition techniques to relaxation-time data from in-vivo MR images. Multiple-point relaxation-time maps were classified using both statistical measures and texture descriptors derived from spatial grey level dependency matrices. These parameters were combined in a new pattern-recognition method based upon the Karhunen- Loeve expansion.