A novel distributed deep learning approach for large-scale chest X-ray covid-19 images detection

Recently, the high mortality rates caused by the covid-19 pandemic have formed a crucial problem to support its enormous and rapid spread. Due to the need for accurate and rapid diagnosis of patients, it is necessary to develop a self-operating recognition model as a fast recognition system to detect covid- 19 infection and prevent the rapid spread among people. Chest X-ray images have been demonstrated to be useful in detecting the covid-19 cases. Further- more, covid-19 being a new disease, the annotated large-scale CXR dataset (big data) for this particular disease is virtually nonexistent. We propose exploiting a large-scale chest X-ray dataset and training a distributed deep neural network to address the limited data issue. In this paper, we propose a new distributed deep learning-based approach (COV-DCNN) for detecting large-scale chest X-ray covid-19 images using ImageDataGenerator to do dynamic data preprocessing and splitting and spark TensorFlow distributor framework to execute distributed training tasks using a Spark job in barrier mode. Furthermore, we adopted a distributed Deep Convolutional Neural Network (DCNN), ANN, AlexNet on Spark Tensor- Flow Distributor to classify large-scale chest X-ray covid-19 images. The highest accuracy was obtained by the DCNN model, which is 91.88% during 0.006s for each case, 84.62% for the ANN model, and 73.8% for the AlexNet model. The obtained results achieve high classification accuracy at a faster speed than other state-of-the-art classification methods