key: cord-0914261-w81sflnr authors: Ameer, Aseel Qassim Abdul; Mohammed, Raghad Falih title: Covid-19 Detection Using CT Scan based on Gray Level Co-Occurrence Matrix date: 2021-04-27 journal: Mater Today Proc DOI: 10.1016/j.matpr.2021.04.224 sha: 8dd144d1b109b135813d660cca374fdc7bcc5b51 doc_id: 914261 cord_uid: w81sflnr The Coronavirus pandemic is one of the biggest problems the world has faced in the 21st century and this virus is a virus that infects the lung and causes breathing problems. In this research the program is designed for the purpose of reading images of the type CT scan, this study used 654 case these cases split in to two classes (infect , not infect), there are two phases in this study, training phase and testing phase. After training the training data store in database, the second phase is testing at first is pre-processing step which increase contrast, then remove lung by labelling the most contrast connected pixels and subtract labelling pixels from original image, the next step is noise removal by applying three filters (mean, median, Gaussian), after that applying gray level co-occurrence matrix (GLCM) in four directions (0°,45°,90° and 135°), then extract features from GLCM, in this study 10 features was extracted from each GLCM matrix, then compare between testing features and training database to specify the case is infect or not, in this study get accuracy 94% for detect the location of infection and detect the lung is infect or not. In Dec. 2019, a disease outbreak has been reported in Wuhan, China, with a new coronavirus that has been referred to as the Severe Acute Respiratory Syndrome Coronavirus II (SARS-CoV2) [1] . The World Health Organization (WHO) after that named this disease as the Coronavirus Disease 2019 (COVID-19). On Mar. 11, 2020, WHO have announced the COVID-19 as pandemic [2] , with 10021401 confirmed cases and 499913 reported deaths all over the world, by Jun. 29, 2020 [3] . The computer vision systems' development has supported medical applications like the increase in image quality, organ texture classification and organ segmentation. The analyses of the tumor properties and time series (2) , detection and segmentation (3) of tumor modules are a few of machine learning applications in the area of the bio-medical image processing [4] . Compared to the chest X-ray, the CT has more clear information and it offers better judgement accuracy hence, this research work considered only the CT for the examination. Rodriguez-Morales et al.27 present a review of the detection and prediction of COVID-19 pneumonia infection [5] . The current procedures used in the identification of the virus require an experienced radiologist, hence automatic detection would be essential to reduce the assessment time for radiologists. The work in 28 reviews the recent image processing techniques. In this research, 654 CT images have been utilized for the classification of COVID-19. Prior to the process of the classification, data-set samples have been classified as coronavirus / non-coronavirus (i.e. infected / non-infected) [6] . Feature extraction approaches with the use of the Euclidian distance and gray level co-occurrence matrix have been utilized throughout the COVID-19 image classification. The results have shown that the suggested approach may be utilized for the diagnosis of COVID-19 as an assisting system [7] . For further studies, refer to [8] , [9] , [18] - [27] , [10] - [17] Dataset used in this study contains 654 CT scan images and is classified as the follows: a. First part contains healthy CT scan image of any infection. This part contains 410 CT scan images. b. The second part contains the CT scan images with non-infection. This part contains 244 images The dataset is split into two parts. The first part consists of 70% images used for training system. The second part consists of 30% images used for testing the proposed system. Figure 1 illustrate CT scan images of dataset. The presented section describes the suggested approach for efficient segmentation and detection of infection from CT scan images which includes five steps (pre-processing, image enhancement, segmentation, features extraction and classification). This step represents the proposed system work as flowchart show in Figure 2 that illustrates the training steps. Figure 3 illustrates the testing steps. The steps of the work are identified in details. The training steps can be split into two parts, the first part is the pre-processing of images, while the second part is applying GLCM and feature extraction. A few steps of pre-processing will be done to the raw input CT scan image; thus, images will be transformed to adequate form for more processing. In such step, the raw image will include a lot of undesired information (noise). Initially a tracking algorithm is used. Step 1: Loading CT scan image Step2: Increase contrast of loading image. Step3: Every one of the elements corresponding to a gray value between (0 and 255). Step 4: Using a 2D matrix for storing the image. Step 5: Lung elimination by read pixels of image and search about first white pixel then label first row of lung, then begin with label every pixel connect with labelling pixels. Step 6: Subtract extracted lung from gray image. Image Enhancement might be defined as a process used to enhance the appearance of images with regard to better visibility and contrast. At this part the images which are 457 image input to preprocessing in order to optimize the CT scan images. This images are classify into types of diseases. The first step in this part is apply three filters which are (mean, Gaussian, and median filters), these filters will removing the noise and enhance the image. In mean filter, a mask size will be determining the loss of details and degree smoothing. Also, the noise that is randomly vary below and over normal brightness value might be decreased via averaging neighborhood of values. Median filter is one of the majorly recognized statistics filters, it is replacing the pixel value by the median regarding grey levels in pixel's neighborhood, while the pixel original value will be involved in computations of median, such filters are widely utilized due to the fact that in specific random noise types, they will be providing efficient capabilities of noisereduction, with significantly less blurring in comparison to the small size linear smoothing filters. Also, such filters are especially important with the existence of unipolar and bipolar impulse noise. Gaussian filter: As soon as calculating an adequate kernel, Gaussian smoothing will be achieved with the use of standard convolution approaches. Actually, the convolution might be fairly quickly achieved as the equation for 2-D isotropic Gaussian was separated into x as well as y components. One of the statistical approaches for examining the textures considering the pixel spatial relations is the GLCM, such approach is characterizing the image texture via estimating how frequently pixel pairs with certain values as well as in certain spatial relationships exists in the image. This achieves the extraction related to the statistical measures from such matrix. GLCM is created via gray-co-matrix function through estimating how regularly a pixel with the intensity (grey-level) value (for example) row happens in certain spatial relationship to the pixel with a value by column and row. The relation is specified as the picture elements in terms of features present and the pixel to adjacent. At this step system will determine shape of extracted region in order to allow to recognize lung contain COVID 19. The first steps is pre-processing which is applying contrast step to increase the light of testing image Then convert image to gray scale Figure (4) illustrates dataset, and the result of contrast and convert to gray scale. Applying three filters to remove noises from the images and enhancement them. Mean filter is the first filter was apply in pre-processing, this filter will remove some types of noises like grain noise from the images, and by using the mask of mean filter will compute average of neighbors pixels and remove noise. B. Applying Median Filter Median filter is most common filter to remove and clean the images from noises, median filter can greatly reducing the time of the cleaning. C. Applying Gaussian Filter: Gaussian filter remove the random variation brightness information from images, by applying Gaussian filter, the types of noises remove are: (Salt and pepper noise, Gaussian noise, and Speckle noise. Then Applying GLCM, the GLCM read texture of tumor in four angles ( , and ) to generate four 0°, 45°90° 135°m atrices for each image. (6). The algorithms and approaches that have been utilized in this system, particularly in the feature extraction stage, are simple and utilize less amount of the memory. It is easy to understand and involves no complicated mathematical formulas. b. Ability of the system can be increased with the use additional characteristics in the input dataset. c. Time consuming for execution of all step in proposed system is less than two second. d. The accuracy of the proposed system reached 94% on a used dataset. f. Determine whether there is a COVID 19 or not as well as in order to facilitate the work of the doctor in the diagnosis and alert if nothing is noticed in the picture. The filters were used to filter CT scan image, which are mostly snouted due to the movement of the patient or the device that was picked up poorly. Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China Added value of computer-aided CT image features for early lung cancer diagnosis with small pulmonary nodules: a matched case-control study Dermatologist-level classification of skin cancer with deep neural networks A deep learning system to screen novel coronavirus disease 2019 pneumonia Classification using link prediction Application of deep learning technique to manage COVID-19 in routine clinical practice using CT images: Results of 10 convolutional neural networks Detection of covid-19 from chest x-ray images using artificial intelligence: An early review Evaluating project management criteria using fuzzy analytic hierarchy Process Evaluating different multimedia learning methodologies by using the MACBETH method Multi-criteria evaluation of e-learning approaches An analytical survey on implementing best practices for introducing e-learning programs to students A comparative analysis for adopting an innovative pedagogical approach of flipped teaching for active classroom learning Selection of suitable e-learning approach using TOPSIS technique with best ranked criteria weights Evaluating of collaborative and competitive learning using MCDM technique An evolutionary algorithm approach for vehicle routing problems with backhauls The effects of perceived conference quality on attendees' behavioural intentions Evaluation of e-learning approaches using AHP-TOPSIS technique Application of a fuzzy multi-objective defuzzification method to solve a transportation problem Analytic hierarchy process for evaluating flipped classroom learning The optimal project selection in portfolio management using fuzzy multi-criteria decision-making methodology Visual sensor intelligent module based image transmission in industrial manufacturing for monitoring and manipulation problems Hyper spectral image classification using dimensionality reduction techniques Efficient cooperative image transmission in one-way multi-hop sensor network Manufacturing intelligent corvus corone module for a secured two way image transmission under WSN Image transmission over decode and forward based cooperative wireless multimedia sensor networks for Rayleigh fading channels in medical Internet of Things (MIoT) for remote health-care and health communication monitoring Effectiveness of artificial intelligence techniques against cyber security risks apply of IT industry Best ways computation intelligent of face cyber attacks