Abstract:With the rapid development of X-ray imaging in the field of medical diagnosis, the traditional method of manual judgment and analysis by doctors based on experience cannot meet the efficiency requirements for diagnosing a large number of chest X-ray images. By improving the residual network for the classification of chest X-ray images and designing an encrypted transmission system, the above-mentioned problem can be effectively solved. The X-ray images are enhanced by Markov random field, and then ResNet50 with strong deep information mining ability is used as the backbone network, the self-attention mechanism is added and the CELU activation function is used to optimize. The experimental test results of Kaggle integrated dataset show that the recall rate of classification is increased from 0.432 to 0.652 while ensuring classification accuracy.Additionally, the system adopts an image encryption algorithm based on logistic chaotic sequences to ensure the privacy of remote medical diagnosis, meeting the application requirements of actual remote medical scenarios.
