To control the spread of the COVID-19 virus and to gain critical time in controlling the spread of the disease, rapid and accurate diagnostic methods based on artificial intelligence are urgently needed. In this article, we propose a clinical decision support system for the early detection of COVID 19 using deep learning based on chest radiographic images. For this we will develop an in-depth learning method which could extract the graphical characteristics of COVID-19 in order to provide a clinical diagnosis before the test of the pathogen. For this, we collected 100 images of cases of COVID-19 confirmed by pathogens, 100 images diagnosed with typical viral pneumonia and 100 images of normal cases. The architecture of the proposed model first goes through a preprocessing of the input images followed by an increase in data. Then the model begins a step to extract the characteristics followed by the learning step. Finally, the model begins a classification and prediction process with a fully connected network formed of several classifiers. Deep learning and classification were carried out using the VGG convolutional neural network. The proposed model achieved an accuracy of 92.5% in internal validation and 87.5% in external validation. For the AUC criterion we obtained a value of 97% in internal validation and 95% in external validation. Regarding the sensitivity criterion, we obtained a value of 92% in internal validation and 87% in external validation. The results obtained by our model in the test phase show that our model is very effective in detecting COVID-19 and can be offered to health communities as a precise, rapid and effective clinical decision support system in COVID-19 detection.

To combat the spread of COVID 19, the World Health Organization suggests a large-scale implementation of COVID 19 tests. Unfortunately, these tests are expensive and cannot be provided and available for people in rural and remote areas. To remedy this problem, we will develop an intelligent clinical decision support system (SADC) for the early diagnosis of COVID 19 from chest x-rays which are more accessible for people in rural areas. Thus, we collected a total of 566 radiological images classified into 3 classes: a class of COVID19 type, a Class of Pneumonia type and a class of Normal type. In the experimental analysis, 70% of the data set was used as training set and 30% was used as the test set. After preprocessing process, we use some augmentation using a rotation, a horizontal flip, a channel shift and rescale. Our finale classifier achieved the best performance with test accuracy of 99%, f1score 98%, precision of 98.60% and sensitivity 98.30%.

In this article, we propose Deep Transfer Learning (DTL) Model for recognizing covid-19 from chest x-ray images. The latter is less expensive, easily accessible to populations in rural and remote areas. In addition, the device for acquiring these images is easy to disinfect, clean and maintain. The main challenge is the lack of labeled training data needed to train convolutional neural networks. To overcome this issue, we propose to leverage Deep Transfer Learning architecture pre-trained on ImageNet dataset and trained Fine-Tuning on a dataset prepared by collecting normal, COVID-19, and other chest pneumonia X-ray images from different available databases. We take the weights of the layers of each network already pre-trained to our model and we only train the last layers of the network on our collected COVID-19 image dataset. In this way, we will ensure a fast and precise convergence of our model despite the small number of COVID-19 images collected. In addition, for improving the accuracy of our global model will only predict at the output the prediction having obtained a maximum score among the predictions of the seven pre-trained CNNs. The proposed model will address a three-class classification problem: COVID-19 class, pneumonia class, and normal class. To show the location of the important regions of the image which strongly participated in the prediction of the considered class, we will use the Gradient Weighted Class Activation Mapping (Grad-CAM) approach. A comparative study was carried out to show the robustness of the prediction of our model compared to the visual prediction of radiologists. The proposed model is more efficient with a test accuracy of 98%, an f1 score of 98.33%, an accuracy of 98.66% and a sensitivity of 98.33% at the time when the prediction by renowned radiologists could not exceed an accuracy of 63.34% with a sensitivity of 70% and an f1 score of 66.67%.