Styliani Petroudi

Prediction of High-Risk Asymptomatic Carotid Plaques Based on Ultrasonic Image Features

Carotid plaques have been associated with ipsilateral neurological symptoms. High-resolution ultrasound can provide information not only on the degree of carotid artery stenosis but also on the characteristics of the arterial wall including the size and consistency of atherosclerotic plaques. The aim of this study is to determine whether the addition of ultrasonic plaque texture features to clinical features in patients with asymptomatic internal carotid artery stenosis (ACS) improves the ability to identify plaques that will produce stroke. 1121 patients with ACS have been scanned with ultrasound and followed for a mean of 4 years. It is shown that the combination of texture features based on second-order statistics spatial gray level dependence matrices (SGLDM) and clinical factors improves stroke prediction (by correctly predicting 89 out of the 108 cases that were symptomatic). Here, the best classification results of 77 ±1.8% were obtained from the use of the SGLDM texture features with support vector machine classifiers. The combination of morphological features with clinical features gave slightly worse classification results of 76 ±2.6%. These findings need to be further validated in additional prospective studies.

Segmentation of the Common Carotid Intima-Media Complex in Ultrasound Images Using Active Contours

The segmentation of the intima-media complex (IMC) of the common carotid artery (CCA) wall is important for the evaluation of the intima media thickness (IMT) on B-mode ultrasound (US) images. The IMT is considered an important marker in the evaluation of the risk for the development of atherosclerosis. The fully automated segmentation algorithm presented in this article is based on active contours and active contours without edges and incorporates anatomical information to achieve accurate segmentation. The level set formulation by Chan and Vese using random initialization provides a segmentation of the CCA US images into different distinct regions, one of which corresponds to the carotid wall region below the lumen and includes the far wall IMC. The segmented regions are used to automatically achieve image normalization, which is followed by speckle removal. The resulting smoothed lumen-intima boundary combined with anatomical information provide an excellent initialization for parametric active contours that provide the final IMC segmentation. The algorithm is extensively evaluated on 100 different cases with ground truth (GT) segmentation available from two expert clinicians. The GT mean IMT value is 0.6679 mm +/ - 0.1350 mm and the corresponding automatically segmented (AS) mean IMT value is 0.6054 mm +/- 0.1464 mm. The mean absolute difference between the GT IMT and the IMT evaluated from from the AS region is 0.095 mm +/ - 0.0615 mm. The polyline distance is 0.096 mm +/ - 0.034 mm while the Hausdorff distance is 0.176 mm +/ - 0.047 mm. The algorithm compares favorably to both automatic and semiautomatic methods presented in the literature.

Quantitative texture analysis of brain white matter lesions derived from T2-weighted MR images in MS patients with clinically isolated syndrome

INTRODUCTION This study investigates the application of texture analysis methods on brain T2-white matter lesions detected with magnetic resonance imaging (MRI) for the prognosis of future disability in subjects diagnosed with clinical isolated syndrome (CIS) of multiple sclerosis (MS). METHODS Brain lesions and normal appearing white matter (NAWM) from 38 symptomatic untreated subjects diagnosed with CIS as well as normal white matter (NWM) from 20 healthy volunteers, were manually segmented, by an experienced MS neurologist, on transverse T2-weighted images obtained from serial brain MR imaging scans (0 and 6-12 months). Additional clinical information in the form of the Expanded Disability Status Scale (EDSS), a scale from 0 to 10, which provides a way of quantifying disability in MS and monitoring the changes over time in the level of disability, were also provided. Shape and most importantly different texture features including GLCM and laws were then extracted for all above regions, after image intensity normalization. RESULTS The findings showed that: (i) there were significant differences for the texture futures extracted between the NAWM and lesions at 0 month and between NAWM and lesions at 6-12 months. However, no significant differences were found for all texture features extracted when comparing lesions temporally at 0 and 6-12 months with the exception of contrast (gray level difference statistics-GLDS) and difference entropy (spatial gray level dependence matrix-SGLDM); (ii) significant differences were found between NWM and NAWM for most of the texture features investigated in this study; (iii) there were significant differences found for the lesion texture features at 0 month for those with EDSS≤2 versus those with EDSS>2 (mean, median, inverse difference moment and sum average) and for the lesion texture features at 6-12 months with EDSS>2 and EDSS≤2 for the texture features (mean, median, entropy and sum average). It should be noted that whilst there were no differences in entropy at time 0 between the two groups, significant change was observed at 6-12 months, relating the corresponding features to the follow-up and disability (EDSS) progression. For the NAWM, significant differences were found between 0 month and 6-12 months with EDSS≤2 (contrast, inverse difference moment), for 6-12 months for EDSS>2 and 0 month with EDSS>2 (difference entropy) and for 6-12 months for EDSS>2 and EDSS≤2 (sum average); (iv) there was no significant difference for NAWM and the lesion texture features (for both 0 and 6-12 months) for subjects with no change in EDSS score versus subjects with increased EDSS score from 2 to 5 years. CONCLUSIONS The findings of this study provide evidence that texture features of T2 MRI brain white matter lesions may have an additional potential role in the clinical evaluation of MRI images in MS and perhaps may provide some prognostic evidence in relation to future disability of patients. However, a larger scale study is needed to establish the application in clinical practice and for computing shape and texture features that may provide information for better and earlier differentiation between normal brain tissue and MS lesions.

An integrated system for the segmentation of atherosclerotic carotid plaque ultrasound video

In this paper, we propose and evaluate an integrated system for the segmentation of atherosclerotic plaque in ultrasound imaging of the carotid artery based on normalization, speckle reduction filtering, and four different snakes segmentation methods. These methods are the Williams and Shah, Balloon, Lai and Chin, and the gradient vector flow (GVF) snake. The performance of the four different plaque snakes segmentation methods was tested on 80 longitudinal ultrasound images of the carotid artery using receiver operating characteristic (ROC) analysis and the manual delineations of an expert. All four methods were very satisfactory and similar in all measures evaluated, with no significant differences between them; however, the Lai and Chin snakes segmentation method gave slightly better results. Concluding, it is proposed that the integrated system investigated in this study could be used successfully for the automated segmentation of the carotid plaque.

A fully automated method using active contours for the evaluation of the intima-media thickness in carotid US images

The thickness of the intima-media complex (IMC) of the common carotid artery (CCA) wall is important in the evaluation of the risk for the development of atherosclerosis. This paper presents a fully automated algorithm for the segmentation of the IMC. The segmentation of the IMC of the CCA wall is important for the evaluation of the intima media thickness (IMT) on B-mode ultrasound images. The presented algorithm is based on active contours and active contours without edges. It begins with image normalization, followed by speckle removal. The level set formulation of Chan and Vese using random initialization provides a segmentation of the CCA ultrasound (US) images into different distinct regions, one of which corresponds to the carotid wall region above the lumen whilst another corresponds to the carotid wall region below the lumen and includes the IMC. The results of the corresponding segmentation combined with anatomical information provide a very accurate outline of the lumen-intima boundary. This outline serves as an excellent initialization for segmentation of the IMC using parametric active contours. The method lends itself to the development of a fully automated method for the delineation of the IMC. The mean and standard deviation of the thickness of the automatically segmented regions are 0.65 mm +/−0.17 mm and the corresponding values for the ground truth IMT are 0.66 mm +/−0.18 mm. The Wilcoxon rank sum test shows no significant difference.

Brain white matter lesion classification in multiple sclerosis subjects for the prognosis of future disability

This study investigates the application of classification methods for the prognosis of future disability on MRI-detectable brain white matter lesions in subjects diagnosed with clinical isolated syndrome CIS of multiple sclerosis MS. In order to achieve these we had collected MS lesions from 38 subjects, manually segmented by an experienced MS neurologist, on transverse T2-weighted images obtained from serial brain MR imaging scans. The patients have been divided into two groups, those belonging to patients with EDSS ≤ 2 and those belonging to patients with EDSS > 2 expanded disability status scale EDSS that was measured at 24 months after the onset of the disease. Several image texture analysis features were extracted from the plaques. Using the Mann-Whitey rank sum test at p 2. These models were based on the Support Vector Machines SVM, the Probabilistic Neural Networks PNN, and the decision trees algorithm C4.5. The highest percentage of correct classifications score achieved was 69% when using the SVM classifier. The findings of this study provide evidence that texture features of MRI-detectable brain white matter lesions may have an additional potential role in the clinical evaluation of MR images in MS.

Segmentation of atherosclerotic carotid plaque in ultrasound video

The degree of stenosis of the common carotid artery (CCA) but also the characteristics of the arterial wall including plaque size, composition and elasticity represent important predictors used in the assessment of the risk for future cardiovascular events. This paper proposes and evaluates an integrated system for the segmentation of atherosclerotic carotid plaque in ultrasound video of the CCA based on normalization, speckle reduction filtering (with the hybrid median filter) and parametric active contours. The algorithm is initialized in the first video frame of the cardiac cycle with human assistance and the moving atherosclerotic plaque borders are tracked and segmented in the subsequent frames. The algorithm is evaluated on 10 real CCA digitized videos from B-mode longitudinal ultrasound segments and is compared with the manual segmentations of an expert, for every 20 frames in a time span of 3-5 seconds, covering in general 2 cardiac cycles. The segmentation results are very satisfactory with a true negative fraction (TNF) of 79.3%, a true-positive fraction (TPF) of 78.12%, a false-positive fraction (FPF) of 6.7% and a false-negative fraction (FNF) of 19.6% between the ground truth and the presented plaque segmentations, a Williams index (KI) of 80.3%, an overlap index of 71.5%, a specificity of 0.88±0.09, a precision of 0.86±0.10 and an effectiveness measure of 0.77±0.09. The results show that integrated system investigated in this study could be successfully used for the automated video segmentation of the carotid plaque.

Combination of different texture features for mammographic breast density classification

Mammographic breast density refers to the prevalence of fibroglandular tissue as it appears on a mammogram. Breast density is not only an important risk for developing breast cancer but can also mask abnormalities. Breast density information can be used for planning individualized screening and treatment. In this work, statistical distributions of different texture descriptors and their combination are investigated with Support Vector Machines (SVMs) for objective breast density classification: Scale Invariant Feature Transforms (SIFT), Local Binary Patterns (LBP) and texton histograms. SIFT is an approach for detecting and extracting local feature descriptors that are reasonably invariant to changes in illumination, image noise, rotation, scaling and small changes in viewpoint. The SIFT descriptor is a coarse descriptor of the edges found in the keypoints. LBPs provide a robust and computationally simple way for describing pure local binary patterns in a texture. They provide information regarding the prevalence of different edge patterns and uniformity. Textons are defined under the operational definition of clustered filter responses and provide a statistical and structural unifying approach for texture characterization. The breast density classification accuracy of the SVM classifiers modeled on the histograms of the three different sets of texture features separately and their combination is evaluated on the Medical Image Analysis Society (MIAS) mammographic database and the results are presented. The combination of the statistical distributions of all the different texture features allows for the highest classification accuracy, reaching over 93%.

Breast density characterization using texton distributions

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 mammograms 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 regions 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.

Segmentation of colorectal pathology images using level sets

Colorectal cancer is the third most common cancer diagnosed in men and women. Generally surgery is by total excision of the mesorectum (TME), though it often has a poor outcome due to affected lymph nodes close to the resection boundary. Advancements in diagnosis and treatment of colorectal cancer require integration of information from different sources such as pathology macroscopic and microscopic images and Magnetic Resonance Images. Evaluation of the mesorectal fascia and the rectal wall are important for both staging the cancer and predicting the outcome of the TME. An algorithm is developed for segmentation of the rectal wall on macroscopic pathology slice images. The information is vital for registration of the images for reconstruction of the resected volume but more importantly for fusion of images in order to evaluate different measures and establish correspondences across modalities. The resected specimen is segmented from the background using thresholding. Following, a number of features such as intensity different order statistics and phase information are evaluated for the region of interest. The features are incorporated in a level set framework for the segmentation of the rectal wall.