Pre-trained VGG-16 with CNN Architecture to classify X-Rays images into Normal or Pneumonia
P Naveen, B. Diwan
Abstract
Chest radiograph (x-ray) are primarily used to diagnose pneumonia However, x-ray examination is a difficult operation, even for a qualified radiologist and there is a need to improve the accuracy of the diagnosis. In this paper, we used chest x-ray images in the Visual Geometry Group (VGG16) concept from Convolutional neural network (CNN) to classify pneumonia or normal that could help doctors in their decision-making phase. In order to refine the training samples in a balanced way, data augmentation methods are used. This method is a supervised learning technique that predicts the outcome, based on the accuracy of the data set used by the network. VGG 16 model were utilized and calibrate the models of deep learning to achieve greater precision in training and validation. Model has successfully implemented in python using Keras library. The model was capable of achieving a 95.67 percent test precision and a 96 AUC score on the dataset with 12.64 test loss and 50 percent accuracy on test.