MLPA-Net: A Novel Lightweight Multi-Channel Attention Convolutional Network for COVID-19 Image Classification Based on Chest X-Ray

Background and Objective: COVID-19 is a new respiratory infectious disease caused by severe acute respiratory syndrome coronavirus type II (SARS-COV-2), which has had a huge impact on the medical system. Because the focus of COVID-19 is in the lungs, doctors are able to determine if a patient has COVID-19 by looking at chest X-rays. This paper uses the advantages of deep learning algorithms for automatic diagnosis and treatment of COVID-19 in lung medical images to automatically detect chest X-ray images.Methods: The multi-channel attention (MLPA) module in the lightweight and multi-channel attention network (MLPA-Net) can extract important features in the complicated image areas, extract the weight of multi-channel, and enrich image feature information. The multilayer perceptron (MLP) in MLPA can randomly assign edge weights and update the weights through backpropagation to obtain more favorable feature information.Results: In the proposed dataset, MLPA-NET achieves 99.1% accuracy and 0.991 F1 score for the two-classification, and 98.4% accuracy and 0.980 F1 score for the three-classification. The effectiveness of the improvements to our method was demonstrated through ablation experiments.Conclusion: MLPA-Net is able to focus on the lesion area of the image, helping doctors to classify efficiently, saving time and improving efficiency. Compared with previously proposed methods, our proposed method hopes to help doctors to classify COVID-19 images more efficiently.