CovFrameNet: An Enhanced Deep Learning Framework for COVID-19 Detection.

The novel coronavirus, also known as COVID-19, is a pandemic that has weighed heavily on the socio-economic affairs of the world. Research into the production of relevant vaccines is progressively being advanced with the development of the Pfizer and BioNTech, AstraZeneca, Moderna, Sputnik V, Janssen, Sinopharm, Valneva, Novavax and Sanofi Pasteur vaccines. There is, however, a need for a computational intelligence solution approach to mediate the process of facilitating quick detection of the disease. Different computational intelligence methods, which comprise natural language processing, knowledge engineering, and deep learning, have been proposed in the literature to tackle the spread of coronavirus disease. More so, the application of deep learning models have demonstrated an impressive performance compared to other methods. This paper aims to advance the application of deep learning and image pre-processing techniques to characterise and detect novel coronavirus infection. Furthermore, the study proposes a framework named CovFrameNet., which consist of a pipelined image pre-processing method and a deep learning model for feature extraction, classification, and performance measurement. The novelty of this study lies in the design of a CNN architecture that incorporates an enhanced image pre-processing mechanism. The National Institutes of Health (NIH) Chest X-Ray dataset and COVID-19 Radiography database were used to evaluate and validate the effectiveness of the proposed deep learning model. Results obtained revealed that the proposed model achieved an accuracy of 0.1, recall/precision of 0.85, F-measure of 0.9, and specificity of 1.0. Thus, the study's outcome showed that a CNN-based method with image pre-processing capability could be adopted for the pre-screening of suspected COVID-19 cases, and the confirmation of RT-PCR-based detected cases of COVID-19.

X-ray Image-Based COVID-19 Patient Detection Using Machine Learning-Based Techniques

In early December 2019, the city of Wuhan, China, reported an outbreak of coronavirus disease (COVID-19), caused by a novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2). On January 30, 2020, the World Health Organization (WHO) declared the outbreak a global pandemic crisis. In the face of the COVID-19 pandemic, the most important step has been the effective diagnosis and monitoring of infected patients. Identifying COVID-19 using Machine Learning (ML) technologies can help the health care unit through assistive diagnostic suggestions, which can reduce the health unit's burden to a certain extent. This paper investigates the possibilities of ML techniques in identifying/ detecting COVID-19 patients including both conventional and exploring from chest X-ray images the effect of viral infection. This approach includes preprocessing, feature extraction, and classification. However, the features are extracted using the Histogram of Oriented (HOG) and Local Binary Pattern (LBP) feature descriptors. Furthermore, for the extracted features classification, six ML models of Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) is used. Experimental results show that the diagnostic accuracy of random forest classifier (RFC) on extracted HOG plusLBP features is as high as 94% followed by SVM at 93%. The sensitivity of the K-nearest neighbour model has reached an accuracy of 88%. Overall, the predicted approach has shown higher classification accuracy and effective diagnostic performance. It is a highly useful tool for clinical practitioners and radiologists to help them in diagnosing and tracking the cases of COVID-19. © 2022 CRL Publishing. All rights reserved.

Face Mask Detection using CNN

Face recognition is an important feature of computer vision. It is used to detect a face and recognize a person and verify the person correctly. Face recognition technology plays an essential role in our everyday lives like in passport checking, smart door, access control, voter verification, criminal investigation, and system to secure public places such as parks, airports, bus stations, and railway stations, etc and many other purposes. While going through the pandemic and the post pandemic situations wearing a mask are compulsory for everyone in order to prevent the transmission of corona virus. This resulted in ineffectiveness of the existing conventional face recognition systems. Hence it is required to improvise the existing systems to get the desired results to detect the masked face at the earliest. This system works in three processes that are image pre-processing, image detection, and image classification. The main aim is to identify that whether a person’s face is covered with mask or not as per the CCTV camera surveillance or a webcam recording. It keeps on checking if a person is wearing mask or not. For classification, feature extraction and detection of the masked faces, Convolutional Neural Network (CNN) and Caffe models are used. These help in easy detection of masked faces with higher accuracy in a very less time and with high security. © 2021 IEEE.

Factors determining generalization in deep learning models for scoring COVID-CT images.

The COVID-19 pandemic has inspired unprecedented data collection and computer vision modelling efforts worldwide, focused on the diagnosis of COVID-19 from medical images. However, these models have found limited, if any, clinical application due in part to unproven generalization to data sets beyond their source training corpus. This study investigates the generalizability of deep learning models using publicly available COVID-19 Computed Tomography data through cross dataset validation. The predictive ability of these models for COVID-19 severity is assessed using an independent dataset that is stratified for COVID-19 lung involvement. Each inter-dataset study is performed using histogram equalization, and contrast limited adaptive histogram equalization with and without a learning Gabor filter. We show that under certain conditions, deep learning models can generalize well to an external dataset with F1 scores up to 86%. The best performing model shows predictive accuracy of between 75% and 96% for lung involvement scoring against an external expertly stratified dataset. From these results we identify key factors promoting deep learning generalization, being primarily the uniform acquisition of training images, and secondly diversity in CT slice position.