Search

BACKGROUND: The coronavirus disease 2019 is a serious and highly contagious disease caused by infection with a newly discovered virus, named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). OBJECTIVE: A Computer Aided Diagnosis (CAD) system to assist physicians to diagnose Covid-19 from chest Computed Tomography (CT) slices is modelled and experimented. METHODS: The lung tissues are segmented using Otsu's thresholding method. The Covid-19 lesions have been annotated as the Regions of Interest (ROIs), which is followed by texture and shape extraction. The obtained features are stored as feature vectors and split into 80:20 train and test sets. To choose the optimal features, Whale Optimization Algorithm (WOA) with Support Vector Machine (SVM) classifier's accuracy is employed. A Multi-Layer Perceptron (MLP) classifier is trained to perform classification with the selected features. RESULTS: Comparative experimentations of the proposed system with existing eight benchmark Machine Learning classifiers using real-time dataset demonstrates that the proposed system with 88.94% accuracy outperforms the benchmark classifier's results. Statistical analysis namely, Friedman test, Mann Whitney U test and Kendall's Rank Correlation Coefficient Test has been performed which indicates that the proposed method has a significant impact on the novel dataset considered. CONCLUSION: The MLP classifier's accuracy without feature selection yielded 80.40%, whereas with feature selection using WOA, it yielded 88.94%. [ABSTRACT FROM AUTHOR]

Classification is a data mining task, which plays a vital role in clinical diagnosis. Irrelevant features will reduce the classifier accuracy. Only a subset of features in a clinical dataset plays a key role in diagnosing the disease. Thus, selecting the relevant features and training the classifier will improve the classifier accuracy. Clinical datasets are subjected to pre-processing, followed by feature selection and classification. In this work, a framework that uses Synergistic firefly algorithm for feature selection and an ensemble classifier for classification has been designed and implemented. The missing values in the clinical datasets are handled using the k-Nearest Neighbour (k-NN) technique. Min–max normalization is used to normalize the data and the normalized data are split into training and testing sets using tenfold cross-validation. Feature selection has been carried out using a wrapper approach that uses the Synergistic firefly algorithm for generating feature subsets and the Levenberg-Marquardt Back Propagation Neural Network to evaluate the performance of the subsets. The ensemble classifier, consisting of the Levenberg-Marquardt Back Propagation Neural Network, Extreme Learning Machine and Naïve Bayes classifier, is trained using the optimal features selected by the Synergistic firefly algorithm. Experimentation has been carried out using eight clinical datasets from the University of California Irvine (UCI) machine learning repository and it has been inferred that selecting the relevant features improves the classifier accuracy. The results obtained prove the efficiency of the decision-making system developed using Synergistic Firefly and the decision-making system can assist the clinician in decision-making to diagnose diseases. [ABSTRACT FROM AUTHOR]

A framework for Computer-Aided Diagnosis (CAD) to diagnose Cavitary TB and Miliary TB from chest Computed Tomography (CT) slices has been designed and implemented. The lung tissues from the CT slices are segmented using region-based Active Contour Model (ACM) and the Region of Interests (ROIs) labelled by an expert radiologist are extracted. Features based on shape and texture are extracted from each ROI. A wrapper-based Improved Artificial Bee Colony Optimization (I-ABCO) algorithm with the accuracy of the Support Vector Machine (SVM) classifier as the fitness function is used to select the optimal subset of features. The search process of I-ABCO is improved using two evaluation functions, namely, rough dependency measure (RDM) and mutual information (MI), to promote better exploitation of the search space. The selected features are used to train the Radial Basis Function Neural Network (RBFNN) classifier, using ten-fold cross validation and the performance is evaluated. Experimentation has been performed on CT slices using two datasets: Tuberculosis dataset and Lung Image Database Consortium-Image Database Resource Initiative (LIDC-IDRI) dataset. The accuracy of the CAD systems using I-ABCO with MI and I-ABCO with RDM for both datasets are (88.34%, 92.63%) and (87.32%, 90.17%), respectively. The CAD system using the basic Artificial Bee Colony Optimization (ABCO) algorithm for feature selection is also experimented with the same datasets and an accuracy of (85.68%, 88%) is achieved. The performance of the proposed algorithms outperforms the ABCO algorithm in diagnostic accuracy, which ensures that selecting the optimal feature subset efficiently has an impact on the classification accuracy. [ABSTRACT FROM AUTHOR]

Segmentation of lung tissue from CT images is a challenging task in computer aided diagnosis systems. In this work, a patient-specific, automated model based approach to segment lung is proposed. Patient specific shape knowledge is obtained by preprocessing the same patient's stack of lung CT slices. The algorithm divides the patient's stack into many groups and the first V slices with maximum lung area from each group are selected and stored in the image database for further processing. Landmark points representing the boundary of the lungs are identified from each slice in the database and stored as shape vectors in the image database. Principal component analysis (PCA) reduces the number of landmark points, retains the major variations in the points in every slice in the database and constructs a shape model called point distribution model (PDM). As the model is generated from the same subject's CT slices, it is initialized at the centroid of the diseased CT slice without manual intervention. The generated shape model can be retained for all the future examinations of the patient. The approach is tested using two datasets: one set with eight tuberculosis CT stacks and the other containing six pneumonia and six lung consolidation CT stacks. The accuracy, average similarity index obtained using the overlap score and dice coefficient for both sets are (97%, 97.9%), (0.970, 0.962) and (0.985, 0.983) respectively. The results show that the approach used in model construction to segment lungs from CT image slices has greatly improved segmentation accuracy. [ABSTRACT FROM AUTHOR]