Enhancing Pneumonia Images: A Comparative Experiment of Image Processing Through 5 Convolutional Neural Networks

Pneumonia is an inflammatory lung disease that mainly affects tiny air sacks such as alveoli in the human body. Its mild symptoms may cause inflammation, cough and fever [1]. The classification and detection of common lung images and pneumonia images are helpful in improving the efficiency of disease diagnosis and treatment. The image detected herein is a chest X-ray. In this paper, the transfer learning method and the visual geometry group (VGG), a kind of convolutional neural network (CNN), are used to detect lung images on a multi-scale and lightweight basis. The model proposed in this paper is applied to the original CXR image dataset, and the data enhancement is conducted. Meanwhile, an automatic pneumonia detection system based on a CNN is created [2]. Traditional pneumonia image classifications are difficult and time-consuming, but in this paper, we solve this problem and process images more quickly and efficiently. After data enhancement, the dataset in this paper can achieve 96% classification accuracy using the pretrained model.