Pneumonia Classification Model by Convolutional Neural Network

Abstract

The purpose of this thesis is to train a model to recognize and classify X-ray images of pneumonia patients from those of normal people by approaches of convolutional neural networks in computer vision. This thesis introduces the types of machine learning algorithms, the advantages of deep learning, the composition of artificial neural networks and their hyperparameters, the structure of convolutional neural networks and the points to be noted during training. This thesis also presents previous relevant studies and researches by other authors on similar topics. Finally, with all the background theory mentioned, TensorFlow, Keras, and Sklearn are used on Google Colab platform to build and train the convolutional neural network. A model with an accuracy of 88% was obtained.

There is still room for improvement in this model. For example, continue training with a larger and more diverse set of input data, build more complex convolutional neural networks, different image enhancement schemes, and various hyperparameter adjustments.