Repository logo
 
Loading...
Profile Picture

Search Results

Now showing 1 - 4 of 4
  • Intelligent support system for CVA diagnosis by cerebral computerized tomography
    Publication . Hajimani, Elmira; Ruano, A. E.; Ruano, M. Graça
    The Cerebral Vascular Accident (CVA) is one of the major causes of death in USA and developed countries, immediately following cardiac diseases and tumors. The increasing number of CVA’s and the requirement of short time diagnosis to minimize morbidity and mortality encourages the development of computer aided diagnosis systems. Early stages of CVA are often undetected by human eye observation of Computer Tomographic (CT) images, thus incorporation of intelligent based techniques on such systems is expected to highly improve their performance. This thesis presents a Radial Basis Functions Neural Network (RBFNN) based diagnosis system for automatic identification of CVA through analysis of CT images. The research hereby reported included construction of a database composed of annotated CT images, supported by a web-based tool for Neuroradiologist registration of his/her normal or abnormal interpretation of each CT image; in case of an abnormal identification the medical doctor was indicted by the software application to designate the lesion type and to identify the abnormal region on each CT’s slice image. Once provided the annotated database each CT image processing considered a pre-processing stage for artefact removal and tilted images’ realignment followed by a feature extraction stage. A large number of features was considered, comprising first and second order pixel intensity statistics as well as symmetry/asymmetry information with respect to the ideal mid-sagittal line of each image. The policy conducted during the intelligent-driven image processing system development included the design of a neural network classifier. The architecture was determined by a Multi Objective Genetic Algorithm (MOGA) where the classifier structure, parameters and image features (input features) were chosen based on the use of different (often conflicting) objectives, ensuring maximization of the classification precision and a good generalization of its performance for unseen data Several scenarios of choosing proper MOGA’s data sets were conducted. The best result was obtained from the scenario where all boundary data points of an enlarged dataset were included in the MOGA training set. Confronted with the NeuroRadiologist annotations, specificity values of 98.01% and sensitivity values of 98.22% were obtained by the computer aided system, at pixel level. These values were achieved when an ensemble of non-dominated models generated by MOGA in the best scenario, was applied to a set of 150 CT slices (1,867,602 pixels). Present results show that the MOGA designed RBFNN classifier achieved better classification results than Support Vector Machines (SVM), despite the huge difference in complexity of the two classifiers. The proposed approach compares also favorably with other similar published solutions, both at lesion level specificity and at the degree of coincidence of marked lesions.
  • An intelligent support system for automatic detection of cerebral vascular accidents from brain CT images
    Publication . Hajimani, Elmira; Ruano, M G; Ruano, Antonio
    Objective: This paper presents a Radial Basis Functions Neural Network (RBFNN) based detection system, for automatic identification of Cerebral Vascular Accidents (CVA) through analysis of Computed Tomographic (CT) images. Methods: For the design of a neural network classifier, a Multi Objective Genetic Algorithm (MOGA) framework is used to determine the architecture of the classifier, its corresponding parameters and input features by maximizing the classification precision, while ensuring generalization. This approach considers a large number of input features, comprising first and second order pixel intensity statistics, as well as symmetry/asymmetry information with respect to the ideal mid-sagittal line. Results: Values of specificity of 98% and sensitivity of 98% were obtained, at pixel level, by an ensemble of non-dominated models generated by MOGA, in a set of 150 CT slices (1,867,602 pixels), marked by a NeuroRadiologist. This approach also compares favorably at a lesion level with three other published solutions, in terms of specificity (86% compared with 84%), degree of coincidence of marked lesions (89% compared with 77%) and classification accuracy rate (96% compared with 88%). (C) 2017 Published by Elsevier Ireland Ltd.
  • A radial basis function classifier for the automatic diagnosis of cerebral vascular accidents
    Publication . Ruano, M. Graça; Hajimani, Elmira; Ruano, Antonio
    A Radial Basis Function Neural Network (RBFNN) based diagnosis system for automatic identification of Cerebral Vascular Accident (CVA) through analysis of Computer Tomographic images (CT) is presented. For the design of a neural network classifier, most published methods just focus on the feature selection aspect and do not consider any approach for determining a model structure that best fits the application at their hand. Moreover, considering the domain of lesion detection from brain tissues, their feature space rarely contains symmetry/asymmetry information with respect to ideal mid-sagittal line. Another issue is how to handle multiple conflicting objectives in the design process, such as the maximization of both specificity and sensitivity, enforcing as well generalization. To deal with these challenges, a Multi Objective Genetic Algorithm (MOGA) based approach is used to determine the architecture of the classifier, its corresponding parameters and input features subject to multiple objectives, as well as their corresponding restrictions and priorities.
  • A software tool for intelligent CVA diagnosis by cerebral computerized tomography
    Publication . Hajimani, Elmira; Ruano, Carina A.; Ruano, M. Graça; Ruano, Antonio
    The final goal of this work is to create an intelligent support system which assists neuroradiologists to identify Cerebral Vascular Accidents in less time, more precisely. For this purpose, the first step was the creation of a web based tool for registering pathological areas in CT images, which will allow to collect required data for training and testing our proposed classifier, a Radial Basis Function (RBF) based Neural Network.