An Efficient ResNet-50 based Intelligent Deep Learning Model to Predict Pneumonia from Medical Images

Abstract

Pneumonia is a devastating disease from which millions of people died every year. Deep learning can be considered as a crucial tool for detecting the disease quickly and accurately. For diagnosis purpose computer assisted methods are less accurate and outdated. So Deep learning is considered as a better option as compared to traditional methods. In this research work, the authors have proposed an efficient ResNet-50 Transfer learning-based Convolutional Neural Network Model to predict pneumonia using medical images. Kaggle based open source dataset repository is used for the experimental analysis. Techniques such as data augmentation, fine tuning, residual blocks, and transfer learning had been used for the better results in terms of maximizing accuracy and minimizing loss. After applying the deep learning methods, the authors achieved an accuracy of 96.6% which is better as compared to other recent literature work. Results shows that this model provided better accuracy and also can be used as a screening test for prediction of pneumonia diagnosis.