Federated Learning for Detecting COVID-19 in Chest CT Images: A Lightweight Federated Learning Approach

The novel coronavirus is spreading rapidly worldwide, and finding an effective and rapid diagnostic method is apriority. Medical data involves patient privacy, and the centralized collection of large amounts of medical data is impossible. Federated learning is a privacy-preserving machine learning paradigm that can be well applied to smart healthcare by coordinating multiple hospitals to perform deep learning training without transmitting data. This paper demonstrates the feasibility of a federated learning approach for detecting COVID-19 through chest CT images. We propose a lightweight federated learning method that normalizes the local training process by globally averaged feature vectors. In the federated training process, the models' parameters do not need to be transmitted, and the local client only uploads the average of the feature vectors of each class. Clients can choose different local models according to their computing capabilities. We performed a comprehensive evaluation using various deep-learning models on COVID-19 chest CT images. The results show that our approach can effectively reduce the communication load of federated learning while having high accuracy for detecting COVID-19 on chest CT images. © 2022 IEEE.

VQ-ViCNet: Strengthen Unique Features Comparison Autoencoder with Embedding Space for Covid-19 Image Classification

In this paper, we propose a new novel coronavirus pneumonia image classification model based on the combination of Transformer and convolutional network(VQ-ViCNet), and present a vector quantization feature enhancement module for the inconspicuous characteristics of lung medical image data. This model extracts the local latent layer features of the image through the convolutional network, and learns the deep global features of the image data through the Transformer's multi-head self attention algorithm. After the calculation of convolution and attention, the features learned by the Transformer Encoder are enhanced by the vector quantization feature enhancement module and able to better complete the final downstream tasks. This model performs better than convolutional architectures, pure attention architectures and generative models on all 6 public datasets. © 2022 IEEE.

Clustering SARS-CoV-2 Variants from Raw High-Throughput Sequencing Reads Data

The massive amount of genomic data appearing over the past two years for SARS-CoV-2 has challenged traditional methods for studying the dynamics of the COVID-19 pandemic. As a result, new methods, such as the Pangolin tool, have appeared which can scale to the millions of samples of SARS-CoV-2 currently available. Such a tool is tailored to take assembled, aligned and curated full-length sequences, such as those provided by GISAID, as input. As high-throughput sequencing technologies continue to advance, such assembly, alignment and curation may become a bottleneck, creating a need for methods which can process raw sequencing reads directly. In this paper, we propose several alignment-free embedding approaches, which can generate a fixed-length feature vector representation directly from the raw sequencing reads, without the need for assembly. Moreover, because such an embedding is a numerical representation, it can be passed to already highly optimized clustering methods such as k-means. We show that the clusterings we obtain with the proposed embeddings are more suited to this setting than the Pangolin tool, based on several internal clustering evaluation metrics. Moreover, we show that a disproportionate number of positions in the spike region of the SARS-CoV-2 genome are informing such clusterings (in terms of information gain), which is consistent with current biological knowledge of SARS-CoV-2. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Increasing validation accuracy of a face mask detection by new deep learning model-based classification

During COVID-19, wearing a mask was globally mandated in various workplaces, departments, and offices. New deep learning convolutional neural network (CNN) based classifications were proposed to increase the validation accuracy of face mask detection. This work introduces a face mask model that is able to recognize whether a person is wearing mask or not. The proposed model has two stages to detect and recognize the face mask;at the first stage, the Haar cascade detector is used to detect the face, while at the second stage, the proposed CNN model is used as a classification model that is built from scratch. The experiment was applied on masked faces (MAFA) dataset with images of 160x160 pixels size and RGB color. The model achieved lower computational complexity and number of layers, while being more reliable compared with other algorithms applied to recognize face masks. The findings reveal that the model's validation accuracy reaches 97.55% to 98.43% at different learning rates and different values of features vector in the dense layer, which represents a neural network layer that is connected deeply of the CNN proposed model training. Finally, the suggested model enhances recognition performance parameters such as precision, recall, and area under the curve (AUC). © 2023 Institute of Advanced Engineering and Science. All rights reserved.

Automatic COVID-19 diagnosis using deep learning features and support vector machines based on Bayesian optimization

Early diagnosis of COVID-19 and detection of infected people are crucial in taking preventative measures and treating the infected people. Artificial intelligence applications based on machine and deep learning techniques are more effective and applicable in such cases. In this work, an approach for automatic COVID-19 diagnosis using chest X-ray images is proposed. In this paper, AlexNet, VGG16, and VGG19 deep learning architectures were used to extract the useful and relevant features. These features were then used as inputs to the support vector machine (SVM) with two discrete outputs: COVID-19 or No-findings. Furthermore, the Bayesian optimization (BO) algorithm was used to tune the parameters of the SVM classifier and choose the optimal parameters. The results of the study indicate that the VGG16-SVM-BO and VGG19-SVM-BO models give the best performance with an accuracy of 99.47%. According to this result, the proposed approach can effectively contribute to the diagnosis of COVID-19. © 2022 IEEE.

AN END-END DEEP LEARNING FRAMEWORK FOR LUNG INFECTION RECOGNITION USING ATTENTION-BASED FEATURES AND CROSS AVERAGE POOLING

Automated detection of lung infections from medical imaging combined with computer vision has a great deal of promise for improving healthcare towards COVID-19 and its consequences due to restricted healthcare emergencies. Finding the affected tissues, segmenting them from lung images is difficult due to comparable neighboring tissues, hazy boundaries, and unpredictable infections. To overcome these issues, we propose a novel deep learning framework that employs attention-based feature vectors and cross average pooling to detect the lung infection from the images. Multimodal images, after enhancement are processed independently through a pretrained DenseNet where the feature extraction is performed from fully connected and average pooled layers. Instead of assigning equal weight to each feature value in the feature vectors, an attention weight is assigned to each feature to highlight how much attention should be paid to it. The attention-based features are then fused using cross average pooling to produce a discriminatory feature set leading to improved diagnosis. The fused features are passed through a deep learning (DL) modified neural network classifier to diagnose the infection. Experiments are performed on the standard Kaggle and Mendeley datasets containing 24,697 X-ray images and 8055 computed topography (CT) images. The results indicated an average accuracy of 99.2%, appreciable Kappa index of 98.11%, and F1 Score of 0.99. A one class accuracy of 99.5% is achieved for COVID-19. The proposed model is robust to noise when tested on degraded images. The results of our DL method for categorizing respiratory tract infections are compared to that of various existing DL models, demonstrating its effectiveness. © 2022 by Begell House, Inc.

A Bag-of-Features (BoF) Based Novel Framework for the Detection of COVID-19

Novel coronavirus (COVID-19) is a hazardous virus. Initially, detected in China and spread worldwide, causing several deaths. Over time, there have been several variants of COVID-19, we have grouped all of them into two major categories. The categories are known to be variants of concern and variants of interest. Talking about the first of these two, it is very dangerous, and we need a system that can not only detect the disease but also classify it without physical interaction with a patient suffering from COVID-19. This paper proposes a Bag-of-Features (BoF) based deep learning framework that can detect as well as classify COVID-19 and all of its variants as well. Initially, the spatial features are extracted with deep convolutional models, while hand-crafted features have been extracted from several hand-crafted descriptors. Both spatial and hand-crafted features are combined to make a feature vector. This feature vector feeds the classifier to classify different variants in respective categories. The experimental results show that the proposed methodology outperforms all the existing methods. © 2021 IEEE.

COVID-19 and Other Lung Disease Detection Using VGG19 Pretrained Features and Support Vector Machine

At present, pandemic phase is declared by World Health Organization caused by COVID-19 disease that endangers all walks of life. The disease has spread quickly around the world causing many countries to lockdown. The medical center could not handle a large number of infected patients. To effectively and automatically classify the infected patients is a big challenge. So, we introduce an efficient lung disease detection method that can detect and identify normal people (without lung disease) and others who have lung disease(s) using chest X-ray images. We consider many well-known lung diseases which are COVID-19, Pneumonia, Pneumothorax, and Atelectasis. First, we preprocessed the images and performed feature extraction using a VGG19 deep-learning model, and then used a Support Vector Machine as the classification model. The dataset that we used is publicly provided with many types of diseases. Our model obtains great results on binary class classification considering COVID-19 and non-COVID-19 classes with 99.0% accuracy, 98.3% recall, 99.1% precision, and 98.7% f1-score. With multi-class classification (5 output classes), we obtain 99.2% accuracy for COVID-19 detection, 99.2% accuracy for Pneumonia, 85.4% accuracy for Atelectasis, and 84.8% accuracy for Pneumothorax. © 2021 IEEE.

Discovering DTI and DDI by Knowledge Graph with MHRW and Improved Neural Network

Drug discovery is of great significance in medical and biological research, while the study of Drug-Target Interaction (DTI) and Drug-Drug Interaction (DDI) can help accelerate drug discovery progress. This paper proposes a new hybrid method for DTI prediction and DDI prediction, which is called MHRW2Vec-TBAN that combines graph representation learning and neural network. MHRW2VecTBAN first constructs knowledge graph KG-DTI and KG-DDI that integrate data related to drugs and targets. Then, an improved graph representation learning model, MHRW2Vec model, is used to obtain feature vectors of reflecting the network structure information for improving the performance of representation learning. Finally, the feature vectors obtained are input to the improved neural network model TextCNN-BiLSTM-Attention Network (TBAN). The experimental results show that, compared with other existing methods, our method could discover more deeper the relationship between drugs and their potential neighborhoods, and has great improvements in DTI prediction and DDI prediction. In addition, the case study of prediction COVID-19 DTI also shows that the proposed model has the potential for actual drug discovery. © 2021 IEEE.