Covid-19 Disease Diagnosis From Radiology Data With Deep Learning Algorithms

Abstract

quick and easy interpretation of the images during the epidemic process that continues all over the world. II. METHODS AND MATERIALS A. Convolutional Neural Network (CNN) ArchitectureCNNs are one of the most commonly used image classification models of neural networks. Corresponding filters in CNN can capture the spatial and time dependence of an image [6]. CNN reduces the image properties to an easier to edit system without reducing the properties required for good classification. CNN's architecture consists of a layer sequence that uses a different function for each layer to transform one layer into another. There are normally three layers to create a CNN model. These are: convolutional layer, pooling layer and fully connected layer. There are two types of pooling in pooling layers. The first type is maximum pooling that returns the maximum value from the core part of the image, eliminating noisy activation and reducing both noise and size. The second type is average pooling that reduces the size of the matrices and uses this reduction as a noise control mechanism. As a result, maximum pooling is better than average pooling. Generally, CNN uses n rows, m columns, and 3 color channels in an image matrix (R, G and B) and the input 3rd order tensor which takes into account the spatial structural structure of the image. The input passes through the convolution layer, the pool layer and the fully connected layer, where the output of each layer is used as input for the following layer. The network begins with 3 * m * n input neurons that are used to encode pixel densities for input features of a 2dimensional image. This is followed by a convolution layer in the local receptor domain fxf. The result, when using a 1-period step 3 × (m-f + 1) × (n-f + 1) it is a layer of latent property neurons. The pooling layer will be applied to 2x2 regions on each of the 3 feature maps and 3 × (m-3 × (mr + 1) / 2 × (nr + 1) / 2 hidden feature neurons will be obtained. The feature map is usually convolution processing by multiplying the Entry matrix element with the filter or core elements, and then the result is collected to obtain a pixel from the feature map.

Performance Assessment of DL Model in Iraq for Covid-19 Patient Severity Prediction from X-Ray Scan Images

“…Mertyuz et al [6] They [8] built a transfer learning pipeline to categorize covid-19 chest X-ray pictures from two datasets of chest X-rays that are available to the public. They obtained an overall detection accuracy of 90%, 94.3%, and 96.8% for the VGG16, ResNet50, and EfficientNetB0 backbones, respectively, using multiple pre-trained convolutional backbones as the feature extractor.…”

“…"Ghost feature maps" are the product of the authors ' excessive repetition. The new [6] 95.87% Zebin, T.,els [8] 90% Brunese, L.,els [9] 97% Improved VGG16 98.8%…”

“…As a result, deep learning is frequently used in diagnostics. [6] X-ray and CT scans are often used along with CAD (Computer Aided Diagnosis) in order to predict Covid-19, as manual work on X-ray and CT images is a very difficult task [7]. A chest CT scan is very useful because it detects cystic fibrosis, lung problems, and chronic lung diseases [6].…”

“…[6] X-ray and CT scans are often used along with CAD (Computer Aided Diagnosis) in order to predict Covid-19, as manual work on X-ray and CT images is a very difficult task [7]. A chest CT scan is very useful because it detects cystic fibrosis, lung problems, and chronic lung diseases [6]. It has been proven that anyone can get Covid19, regardless of gender, age or race.…”