An Automated CAD System of CT Chest Images for COVID-19 Based on Genetic Algorithm and K-Nearest Neighbor Classifier

Abstract

The detection of COVID-19 from computed tomography (CT) scans suffered from inaccuracies due to its difficulty in data acquisition and radiologist errors. Therefore, a fully automated computer-aided detection (CAD) system is proposed to detect coronavirus versus non-coronavirus images. In this paper, a total of 200 images for coronavirus and non-coronavirus are employed based on 90% for training images and 10% testing images. The proposed system comprised five stages for organizing the virus prevalence. In the first stage, the images are preprocessed by thresholding-based lung segmentation. Afterward, the feature extraction technique was performed on segmented images, while the genetic algorithm performed on sixty-four extracted features to adopt the superior features. In the final stage, the K-nearest neighbor (KNN) and decision tree are applied for COVID-19 classification. The findings of this paper confirmed that the KNN classifier with K=3 is accomplished for COVID-19 detection with high accuracy of 100% on CT images. However, the decision tree for COVID-19 classification is achieved 95% accuracy. This system is used to facilitate the radiologist’s role in the prediction of COVID-19 images. This system will prove to be valuable to the research community working on automation of COVID-19 images prediction.

Metaheuristics based COVID-19 detection using medical images: A review

“… Gray Level Co-Occurrence Matrix (GLCM) GLDM is mostly used for extracting texture features from images on the basis of correlation of image pixels [ 15 , 22 ] Gray Level Run Length Matrix (GLRLM) GLRLM can extract high order texture features, like short-run emphasis, long-run emphasis, etc. [ 40 ]. SFTA This texture feature extraction technique consists of two steps, creating image stacks and analysis of each binary image based on boundaries, pixel count, mean gray level [ 67 ].…”

“…This algorithm is categorized in the following main steps: initialize population, create initial solutions (here, any function according to the problem can be defined), select the best members, perform crossover and mutation, and then select members for the next generation. These steps should be repeated until the desired population is created [ 40 , 67 , 69 ]. Fig.…”

“…The main drawback of KNN is its high memory consumption and its sensitivity toward outliers. Few studies by Afify et al [ 40 ] and Shaban et al [ 69 ] have used it as the last step of the classification pipeline (COVID-19 classification), and also adopted nature-inspired algorithm GA for feature selection. One study [ 69 ] tried to reduce the drawback of sensitivity toward outliers by introducing a heuristic-based approach to decide.…”

Fusion of Moment Invariant Method and Deep Learning Algorithm for COVID-19 Classification

“…The most important factor in the success rate of machine learning is the amount of training provided and the availability of testing data [39]. According to experts' opinions, test outcomes will fail to attain a good classifier if training data is below 50% [40]. Hence, to be assured of achieving an accurate diagnosis, the training data were increased.…”

Analysis of Classroom Learning Behaviors Based on Internet of Things and Image Processing

“…Chao et al [10] extracted the differential entropies of EEG signals, trained deep belief network (DBN) and DBNhidden Markov model (HMM), and achieved high classification accuracy of two types of emotions. Aditya and Tibarewala [11] preprocessed the EEG signals from ten channels through discrete wavelet transform (DWT), extracted energy entropy as features, and classified them with support vector machine (SVM) and k-nearest neighbors (KNN) algorithm; experimental results show that band was classified more accurately than the other low-frequency channels [12][13][14][15][16].…”