PSCNN: PatchShuffle Convolutional Neural Network for COVID-19 Explainable Diagnosis

Audio texture analysis of COVID-19 cough, breath, and speech sounds

The coronavirus disease (COVID-19) first appeared at the end of December 2019 and is still spreading in most countries. To diagnose COVID-19 using reverse transcription - Polymerase chain reaction (RT-PCR), one has to go to a dedicated center, which requires significant cost and human resources. Hence, there is a requirement for a remote monitoring tool that can perform the preliminary screening of COVID-19. In this paper, we propose that a detailed audio texture analysis of COVID-19 sounds may help in performing the initial screening of COVID-19. The texture analysis is done on three different signal modalities of COVID-19, i.e. cough, breath, and speech signals. In this work, we have used 1141 samples of cough signals, 392 samples of breath signals, and 893 samples of speech signals. To analyze the audio textural behavior of COVID-19 sounds, the local binary patterns LBP) and Haralick’s features were extracted from the spectrogram of the signals. The textural analysis on cough and breath sounds was done on the following 5 classes for the first time: COVID-19 positive with cough, COVID-19 positive without cough, healthy person with cough, healthy person without cough, and an asthmatic cough. For speech sounds there were only two classes: COVID-19 positive, and COVID-19 negative. During experiments, 71.7% of the cough samples and 72.2% of breath samples were classified into 5 classes. Also, 79.7% of speech samples are classified into 2 classes. The highest accuracy rate of 98.9% was obtained when binary classification between COVID-19 cough and non-COVID-19 cough was done.

A facial expression recognizer using modified ResNet-152

In this age of artificial intelligence, facial expression recognition is an essential pool to describe emotion and psychology. In recent studies, many researchers have not achieved satisfactory results. This paper proposed an expression recognition system based on ResNet-152. Statistical analysis showed our method achieved 96.44% accuracy. Comparative experiments show that the model is better than mainstream models. In addition, we briefly described the application of facial expression recognition technology in the IoT (Internet of things).

A Review of Image Classification Algorithms in IoT

With the advent of big data era and the enhancement of computing power, Deep Learning has swept the world. Based on Convolutional Neural Network (CNN) image classification technique broke the restriction of classical image classification methods, becoming the dominant algorithm of image classification. How to use CNN for image classification has turned into a hot spot. After systematically studying convolutional neural network and in-depth research of the application of CNN in computer vision, this research briefly introduces the mainstream structural models, strengths and shortcomings, time/space complexity, challenges that may be suffered during model training and associated solutions for image classification. This research also compares and analyzes the differences between different methods and their performance on commonly used data sets. Finally, the shortcomings of Deep Learning methods in image classification and possible future research directions are discussed.

COVID-19 CT image recognition algorithm based on transformer and CNN

Novel corona virus pneumonia (COVID-19) broke out in 2019, which had a great impact on the development of world economy and people's lives. As a new mainstream image processing method, deep learning network has been constructed to extract medical features from chest CT images, and has been used as a new detection method in clinical practice. However, due to the medical characteristics of COVID-19 CT images, the lesions are widely distributed and have many local features. Therefore, it is difficult to diagnose directly by using the existing deep learning model. According to the medical features of CT images in COVID-19, a parallel bi-branch model (Trans-CNN Net) based on Transformer module and Convolutional Neural Network module is proposed by making full use of the local feature extraction capability of Convolutional Neural Network and the global feature extraction advantage of Transformer. According to the principle of cross-fusion, a bi-directional feature fusion structure is designed, in which features extracted from two branches are fused bi-directionally, and the parallel structures of branches are fused by a feature fusion module, forming a model that can extract features of different scales. To verify the effect of network classification, the classification accuracy on COVIDx-CT dataset is 96.7%, which is obviously higher than that of typical CNN network (ResNet-152) (95.2%) and Transformer network (Deit-B) (75.8%). These results demonstrate accuracy is improved. This model also provides a new method for the diagnosis of COVID-19, and through the combination of deep learning and medical imaging, it promotes the development of real-time diagnosis of lung diseases caused by COVID-19 infection, which is helpful for reliable and rapid diagnosis, thus saving precious lives.

A Deep Learning-Based Quantitative Structure-Activity Relationship System Construct Prediction Model of Agonist and Antagonist with High Performance

Molecular design and evaluation for drug development and chemical safety assessment have been advanced by quantitative structure–activity relationship (QSAR) using artificial intelligence techniques, such as deep learning (DL). Previously, we have reported the high performance of prediction models molecular initiation events (MIEs) on the adverse toxicological outcome using a DL-based QSAR method, called DeepSnap-DL. This method can extract feature values from images generated on a three-dimensional (3D)-chemical structure as a novel QSAR analytical system. However, there is room for improvement of this system’s time-consumption. Therefore, in this study, we constructed an improved DeepSnap-DL system by combining the processes of generating an image from a 3D-chemical structure, DL using the image as input data, and statistical calculation of prediction-performance. Consequently, we obtained that the three prediction models of agonists or antagonists of MIEs achieved high prediction-performance by optimizing the parameters of DeepSnap, such as the angle used in the depiction of the image of a 3D-chemical structure, data-split, and hyperparameters in DL. The improved DeepSnap-DL system will be a powerful tool for computer-aided molecular design as a novel QSAR system.

COV-RadNet: A Deep Convolutional Neural Network for Automatic Detection of COVID-19 from Chest X-Rays and CT Scans

With the increase in severity of COVID-19 pandemic situation, the world is facing a critical fight to cope up with the impacts on human health, education and economy. The ongoing battle with the novel corona virus, is showing much priority to diagnose and provide rapid treatment to the patients. The rapid growth of COVID-19 has broken the healthcare system of the affected countries, creating a shortage in ICUs, test kits, ventilation support system. etc. This paper aims at finding an automatic COVID-19 detection approach which will assist the medical practitioners to diagnose the disease quickly and effectively. In this paper, a deep convolutional neural network, 'COV-RadNet' is proposed to detect COVID positive, viral pneumonia, lung opacity and normal, healthy people by analyzing their Chest Radiographic (X-ray and CT scans) images. Data augmentation technique is applied to balance the dataset 'COVID 19 Radiography Dataset' to make the classifier more robust to the classification task. We have applied transfer learning approach using four deep learning based models: VGG16, VGG19, ResNet152 and ResNext 101 to detect COVID-19 from chest X-ray images. We have achieved 97% classification accuracy using our proposed COV-RadNet model for COVID/Viral Pneumonia/Lungs Opacity/Normal, 99.5% accuracy to detect COVID/Viral Pneumonia/Normal and 99.72% accuracy to detect COVID and non-COVID people. Using chest CT scan images, we have found 99.25% accuracy to classify between COVID and non-COVID classes. Among the performance of the pre-trained models, ResNext 101 has shown the highest accuracy of 98.5% for multiclass classification (COVID, viral pneumonia, Lungs opacity and normal).

A lightweight double-channel depthwise separable convolutional neural network for multimodal fusion gait recognition

Gait recognition is an emerging biometric technology that can be used to protect the privacy of wearable device owners. To improve the performance of the existing gait recognition method based on wearable devices and to reduce the memory size of the model and increase its robustness, a new identification method based on multimodal fusion of gait cycle data is proposed. In addition, to preserve the time-dependence and correlation of the data, we convert the time-series data into two-dimensional images using the Gramian angular field (GAF) algorithm. To address the problem of high model complexity in existing methods, we propose a lightweight double-channel depthwise separable convolutional neural network (DC-DSCNN) model for gait recognition for wearable devices. Specifically, the time series data of gait cycles and GAF images are first transferred to the upper and lower layers of the DC-DSCNN model. The gait features are then extracted with a three-layer depthwise separable convolutional neural network (DSCNN) module. Next, the extracted features are transferred to a softmax classifier to implement gait recognition. To evaluate the performance of the proposed method, the gait dataset of 24 subjects were collected. Experimental results show that the recognition accuracy of the DC-DSCNN algorithm is 99.58%, and the memory usage of the model is only 972 KB, which verifies that the proposed method can enable gait recognition for wearable devices with lower power consumption and higher real-time performance.