Machine Learning Algorithm for Trend Analysis in Short term Forecasting of COVID-19 using Lung X-ray Images

With the development of medical technology, the diagnosis of lung diseases relies more on the determination of medical images. With increasingly huge data, a powerful data processing model is urgently needed to provide favorable support for this field. The goal of this study is to develop a computer-assisted method to identify COVID-19 from X-ray pictures of the lungs at the very beginning of the disease. The architecture is implemented as a software system on a computer that can assist in the affordable and accurate early identification of cardiac illness. The performance of CNN architecture is best among all other classification algorithms to detect COVID-9 from Lung X-ray images. The datasets consist of COVID-19 established cases for 4 weeks which included the X-ray images of the chest. Then the distribution of the data was examined according to the statistical distribution. For this prediction, time series models are used for forecasting the pandemic situation. The performances of the methods were compared according to the MSE metric and it was seen that the Convolutional Neural Networks (CNN) achieved the optimal trend pattern.

Lung X-Ray Images Preprocessing Algorithms for COVID-19 Diagnosing Intelligent Systems

Most of the current COVID-19 smart diagnostic systems use convolutional neural networks to detect foci and classify the stages of the disease on X-rays of the lungs. The conditions for obtaining such images and their parameters vary considerably. Therefore, the direct application of convolutional neural networks to unprepared lung X-rays is characterized by low accuracy. To solve this problem, the paper describes models and algorithms for preprocessing lung X-rays in relation to cleaning, preliminary detection in areas of interest, format transformation and standardization of image presentation. The use of such algorithms can significantly improve the accuracy of the subsequent application of convolutional neural networks to detect foci and classify the stages of diseases on prepared X-ray images of the patient's lungs. The paper presents the results of theoretical and experimental studies on real data - on the open depersonalized Kaggle dataset. Based on the presented research results, a generalized model and practical recommendations are provided on the use of particular preprocessing algorithms for the effective diagnosis of COVID19 on heterogeneous chest X-rays. © 2022 Copyright for this paper by its authors.

An Artificial Intelligence Technique for Covid-19 Detection with eXplainability using Lungs X-Ray Images

According to the World Health Organization, the coronavirus outbreak poses a daily threat to the global health system. Almost all countries' health resources are insufficient or unequally distributed. There are several issues, such as a lack of health care workers, beds, and intensive care units, to name a few. The key to the country's health systems overcoming this epidemic is to use limited resources at optimal levels. Disease detection is critical to averting an epidemic. The greater the success, the more tightly the covid viral spread may be managed. PCR (Polymerase chain reaction) testing is commonly used to determine whether or not a person has a virus. Deep learning approaches can be used to classify chest X-RAY images in addition to the PCR method. By analyzing multi-layered pictures in one go and establishing manually entered parameters in machine learning, deep learning approaches have become prominent in academic research. This popularity has a favorable impact on the available health datasets. The goal of this study was to detect disease in persons who had x-rays done for suspected COVID-19 (Coronavirus Disease-2019). A bi-nary categorization has been used in most COVID-19 investigations. Chest x-rays of COVID-19 patients, viral pneumonia patients, and healthy patients were obtained from IEEE [17] (Institute of Electrical and Electronics Engineers) and Kaggle [18]. Before the classification procedure, the data set was subjected to a data augmentation approach. These three groups have been classified through multiclassclassification deep learning models. We are also debating a taxonomy of recent contributions on the eXplainability of Artificial Intelligence (XAI). © 2022 IEEE.

Classification of Covid Patients Based on Detection of Lung X-Rays Using Local Binary Pattern Method

High number of deaths due to Covid-19 outbreak affect people in various ways including their economic and psychological side. Previous studies were carried out in analyzing various symptoms in COVID-19 patients. Patients in severe conditions are usually found with a white spot in their lungs. Therefore chest x-ray is one of the necessary medical assessment to examine the patients. This study focus on determining whether a patient suffered from COVID-19 by analyzing their chest X-rays photos. A total of 864 X-rays photos were used as a dataset. Earlier steps in processing the dataset included removing the noise, equalizing the size and increasing the accuracy value. The Local Binary Pattern (LBP) method was used to extract the dataset feature. The performance analysis result was a precision value of 78.5%, recall of 78%, and f-measure of 79%. © 2021 IEEE.

A Feature Embedding and Multi-dilation Convolutional Neural Network for Automatic COVID-19 Lesion Detection from Lung X-ray Images

Accurate detection of COVID-19 has become one of the major challenges since the outbreak of the pandemic. An effective and rapid detection will prevent the further spread of the deadly disease and enable doctors to treat infected patients appropriately. Therefore, this paper proposes an architecture to detect COVID-19 lesion areas in lung X-ray images based on Cascade-CNN and is named as COVID-Cascade RCNN. This algorithm integrates multi-scale dilated convolution and gender characteristic data from embedding patients. Firstly, to tackle the problem posed by the different size distribution of lesions in lung X-ray images, the basic Resnet101 network is adopted to perform preliminary feature extraction on the detection images. In addition, a multi-dilation convolutional neural network is added to generate more effective multi-scale featured information. The parallel dilated convolution with different dilation rates can extract more local feature information lesion areas of various sizes, improving the model's adaptability and accuracy. Secondly, gender characteristics that generate distinct effects on COVID-19 infection were innovatively embedded into the network model to improve detection accuracy further. Finally, experimental results on public dataset BIMCV-COVID19 show that compared with the previous model, the reformed one significantly improved detection precision with an average precision mAP of 42.197%, which is 5.368% higher than the original model. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.