Deep Learning Chest X-Rays of Pneumonia Binary Classification Based on Convolutional Neural Network for IoT networking

Pneumonia is a severe lung infection caused by a variety of viruses. Due to its similarities to other lung disorders, pneumonia can sometimes be challenging to recognize and address on chest X-ray pictures. This means that no currently available techniques for forecasting pneumonia can achieve sufficiently high levels of accuracy. In order to streamline the operation of identifying pneumonia from chest X-ray pictures online using Internet of Things (IoT) networks, this research introduces a classification method used a custom-made version of the Convolutional Neural Network (CNN) model. We propose to use a structure of three parts: the input to the first part will be split into two identical paths, where each path constructed from four-layers (one of them is Convolution-layer), but with different initialization parameters and distinct dropout values. the results will be added to each other. Then the second part starts with identical two paths again and summed at their last layer. Finally, the third part will be constructed from nine-layers (first two-layers are convolutional-layers). Thanks to kaggle, the model was trained using Kaggel Tensorflow Processor Units (TPUs), which is a free facility from kaggel. Therefore, the training time was dramatically low. Results show that the suggested model achieved precision, recall, and accuracy of 98.94%, 98.55%, and 98.14%, respectively. Thus, the collected results prove that this network can be easily integrated with IoT networks.