Pneumonia Detection Using Deep Learning Techniques

During this time of COVID-19, pneumonia needs to be diagnosed and treated properly. An accurate diagnosis of pneumonia requires an analysis of chest X-rays (CXRs). This requires the expertise of a radiologist. The fuzziness of CXR images makes diagnosis quite challenging and time-consuming, even for experienced radiologists. Human judgment can also lead to erroneous identification. The system can therefore play a significant role here if it is authentic and automated. Every sector uses Deep Learning (DL) in this era of cutting-edge technology. Detecting diseases affecting the chest is possible through chest radiography. Medical imaging research is incorporating computer vision techniques to detect disease. The literature on pneumonia detection is summarised and guidelines for medical practitioners are lacking, despite the fact that several algorithms exist for its detection. In this paper, CNN(Convolution Neural Networks) and VGG-16(Visual Geometry Group) are the algorithms used. The accuracy obtained when training CNN is 89% and for VGG-16 is 85%. The Adam Optimizer is used to assist in improving the accuracy and reduce the total loss. The precision and recall for CNN is 0.79, 0.81 and for VGG-16 is 0.83, 0.81. The Dataset is taken from the Kaggle with the size of 2GB containing 16,000 images of chest Xrays.