Textural analysis of contrast‐enhanced MR images of the breast

Abstract

Texture analysis was applied to high-resolution, contrast-enhanced (CE) images of the breast to provide a method of lesion discrimination. Significant differences were seen between benign and malignant lesions for a number of textural features, including entropy and sum entropy. Using logistic regression analysis (LRA), a diagnostic accuracy of A z ‫؍‬ 0. 80 Key words: texture analysis; breast imaging; contrast enhancement; co-occurrence matrices; image processing A fundamental goal of any diagnostic imaging investigation is tissue classification. Contrast-enhanced (CE) MRI of the breast has excellent sensitivity (1-3), but the reported specificity has varied greatly due to the rapid enhancement of fibroadenomas and some radial scars, in addition to the swift contrast uptake evident in malignant tumors. Various methods to improve specificity have been suggested, such as pharmacokinetic modeling of the time course of contrast enhancement (4,5). Other investigations have concentrated on refining methods of empirical analysis (6,7). However, these methods either aggregate data from the whole lesion or limit data sampling to specific areas of the lesion, and as such disregard morphological and textural information.Until recently, little attention had been paid to the diagnostic potential of texture analysis of MR images. When radiologists make a diagnosis, they employ the local properties of the tissue of interest in their assessment. Although visual assessment of texture is highly subjective, it is known to be a particularly sensitive means of determining pathology (8). Texture analysis is an attempt to quantify and emulate the expert eye of the radiologist. Unfortunately, the texture of an image is an ill-defined property, and texture analysis is a generic name for a series of techniques used to quantify spatial variation of the gray tones in an image. Various textural algorithms have been proposed, including the spatial gray-level dependence method, fractal methods, and Markov random field models (9). The most commonly used statistical technique is the spatial gray-level dependence matrix method proposed by Haralick et al. (10), because of its ability to study the second-order statistics of pixels at different spacings and angles. In comparative studies this method was found to provide as much discriminatory power as other techniques (11), and it performs well for small regions.Texture analysis has previously been applied to a limited extent in MRI. Kjaer et al. (12) reported good discrimination between normal brain tissues, but found texture analysis to be less successful in discriminating between benign and malignant tumor growth. Other such work in the brain has been published (8,13), and a classification rate of 91% was detailed in a study of Alzheimer's disease (14). Applications have also been reported in MR studies of prostate (15), skeletal muscle (16), and bone (17).With regard to breast disease, it has been suggested that ultrasonic image texture analysis could be used to reduce the number of benign l...

Texture Analysis: A Review of Neurologic MR Imaging Applications

“…We can also construct a logistic regression model by assigning the most discriminating features as predictors and either tissue class or diagnosis as the outcome measure. 31 The classification accuracy is then calculated by measuring the area under the ROC curve. We should also note that unsupervised classification techniques including K-means or hierarchical clustering 32 are suitable for scenarios in which there is no prior knowledge of how the feature space is organized; therefore, all cases belong to a single testing set.…”

Mass classification in breast DCE-MR images using an artificial neural network trained via a bee colony optimization algorithm

“…Texture analysis approaches can be summarized into four texture modelling methods: statistical methods, geometrical methods, model based methods, and signal processing methods 39 . The statistical textures are found to be the best for image classification 33 . A statistical approach sees an image texture as a quantitative measure of the arrangement of intensities in a region.…”

“…A typical breast DCE-MRI contains a great amount of heterogeneous information that depicts different tissues, vessels, ducts, chest skin, and breast edge characteristics. Texture features have been widely used in breast DCE-MRI mass classification 7,18,[32][33][34][35][36][37] . The implemented feature extraction procedure relies on the exploration of the textural characteristics of the extracted mass.…”

Comparison of Multi-resolution Analysis Patterns for Texture Classification of Breast Tumors Based on DCE-MRI

“…Dynamic features [3,4] have been used to characterize the temporal enhancement pattern of a tumor, while architectural features [3,4] have been extracted to characterize the morphology of the tumor. Moreover, kinetic [5,6] and texture features [7,8] have been used to distinguish between malignant and benign tumors. More specifically, Yao et al [8] computed textural features based on the cooccurrence matrix and also extracted frequency features by applying the discrete wavelet transform (DWT) on the texture temporal sequences of the breast tumors in order to classify them.…”