Statistical approach is a valuable way to describe texture primitives. The aim of this study is to design and implement a classifier framework to automatically identify the thyroid nodules from ultrasound images. Using rigorous mathematical foundations, this article focuses on developing a discriminative texture analysis method based on texture variations corresponding to four biological areas (normal thyroid, thyroid nodule, subcutaneous tissues, and trachea). Our research follows three steps: automatic extraction of the most discriminative first-order statistical texture features, building a classifier that automatically optimizes and selects the valuable features, and correlating significant texture parameters with the four biological areas of interest based on pixel classification and location characteristics. Twenty ultrasound images of normal thyroid and 20 that present thyroid nodules were used. The analysis involves both the whole thyroid ultrasound images and the region of interests (ROIs). The proposed system and the classification results are validated using the receiver operating characteristics which give a better overall view of the classification performance of methods. It is found that the proposed approach is capable of identifying thyroid nodules with a correct classification rate of 83 % when whole image is analyzed and with a percent of 91 % when the ROIs are analyzed.

Liver fibrosis accurate staging is vital to define the state of the Schistosomiasis disease for further treatment. The present work analyzed the microscopic liver images to identify and to differentiate between healthy, cellular, fibrocellular, and fibrous liver pathologies by proposing a fast, robust, and highly discriminative method based on texture analysis. The multiclass classification based on the "one-versus- all" method that built a voting rule approach to classify the liver images based on the liver state. Specifically, quantitative parameters, such as the anisotropy and laminarity are proposed based on the relative orientation of the pixel pairs in a global and local coherence of gradient vectors approach. Analysis of the tissue texture data using both gradient vector and gradient angle co-occurrence matrix approaches facilitated more definitive identification of the abnormal tissue. The experimental results established that the local anisotropy based texture measures are appropriated for the microtexture analysis in order to discriminate between pathologies. Macrotexture description using the global features provided only integral anisotropy coefficient that has a confidence level similar to those provided by the local feature.

This paper proposes a new hybrid approach to estimate the cardiac cycle phases in 2-D echocardiographic images as a first step in cardiac volume estimation. We focused on analyzing the atrial systole and diastole events by using the geometrical position of the mitral valve and a set of three image features. The proposed algorithm is based on a tandem of image processing methods and artificial neural networks as a classifier to robustly extract anatomical information. An original set of image features is proposed and derived to recognize the cardiac phases. The aforementioned approach is performed in two denoising scenarios. In the first scenario, the images are corrupted with Gaussian noise, and in the second one with Rayleigh noise distribution. Our hybrid algorithm does not involve any manual tracing of the boundaries for segmentation process. The algorithm is implemented as computer-aided diagnosis (CADi) software. A dataset of 150 images that include both normal and infarct cardiac pathologies was used. We reported an accuracy of 90 % and a 2 ± 0.3 s in terms of execution time of CADi application in a cardiac cycle estimation task. The main contribution of this paper is to propose this hybrid method and a set of image features that can be helpful for automatic detection applications without any user intervention. The results of the employed methods are qualitatively and quantitatively compared in terms of efficiency for both scenarios.

No abstract

Abstract. The abnormalities appear frequently on Magnetic Resonance Images (MRI) of brain in elderly patients presenting either stroke or cognitive impairment. Detection of brain hemorrhage lesions in MRI is an important but very time-consuming task. This research aims to develop a method to extract brain tissue features from T2-weighted MR images of the brain using a selection of the most valuable texture features in order to discriminate between normal and affected areas of the brain. Due to textural similarity between normal and affected areas in brain MR images these operation are very challenging. A trauma may cause microstructural changes, which are not necessarily perceptible by visual inspection, but they could be detected by using a texture analysis. The proposed analysis is developed in five steps: i) in the pre-processing step: the de-noising operation is performed using the Daubechies wavelets; ii) the original images were transformed in image features using the first order descriptors; iii) the regions of interest (ROIs) were cropped from images feature following up the axial symmetry properties with respect to the mid -sagittal plan; iv) the variation in the measurement of features was quantified using the two descriptors of the co-occurrence matrix, namely energy and homogeneity; v) finally, the meaningful of the image features is analyzed by using the t-test method. P-value has been applied to the pair of features in order to measure they efficacy.