Evaluation of Deep Learning Models in the Prediction of Lung Disease(Pneumonia)

Medical diagnosis is one of the fields in which there is a great scope for deep neural networks. Pneumonia is a severe infectious disease that can be predicted using chest x-rays by radiologists. Using deep learning, one can develop an automated prediction system by building a convolutional network that can help in predicting the disease. In a deep learning framework, the transfer learning technique is one of the methods based on the image net classifications. In transfer learning, there are many architectures in Keras which are a type of convolution neural networks to build the proposed model which can predict pneumonia disease. The data sets which will be used include several chest x-rays or radiographs which are grouped into normal ones and the ones with pneumonia disease. Prediction of diseases like Pneumonia using such techniques have been in existence for many years but failed to achieve perfect accuracy. Traditional approaches like building a neural network from scratch or other convolutional models have some limitations. In this work, evaluating the effectiveness of the application of transfer learning while comparing various architectures in Pneumonia identification using the chest x-ray data sets is performed. The experimental results revealed that the proposed scheme is effective and shows improved performance over many other existing models.