Optimal Pneumonia detection using Convolutional Neural Networks from X-ray Images

Pneumonia is an inflammatory condition caused by bacterial or viral infection. Individuals having Pneumonia either have one or both lungs affected. Some combinations of productive or dry cough, fever, chest pain and respiratory difficulty include common symptoms of pneumonia. Due to the limited amount of skilled radiologist available particularly in remote areas, there is an urgent need for automatic diagnosis of the disease to help physicians in the treatment process. Due to advancement in computational power of the machine and huge amount of data available, Deep learning techniques are providing promising results. Convolution Neural network is providing efficient solution in medical image classification. In this work, The pretrained model of CNN is used for detection of Pneumonia using chest Xray of normal and Pneumonia affected persons. The pretrained model is finetuned here and compared among three optimizers ie Adam, SGD and RMSProp, and it is inferred that RMSProp Optimizer is providing more efficient model. Here metrics for performance evaluation is accuracy. RMSProp optimizer is giving accuracy of 96.14% compared to other two optimization techniques.