Convolutional Neural Network Model for Analysis of X Rays For Pneumonia Detection
Pneumonia, an interstitial lung sickness as well as life threatening infectious disease, which gained the importance and has become a major killer in the ongoing COVID pandemic. It is also the main source of death in youngsters aged less than five, people above the age of 65 and others with comorbidities in India in accordance with reports published by UNICEF. Ideal identification of pneumonia can assist with early treatment and saving the lives. This paper presents Convolutional Neural Organization models to precisely distinguish pneumonic lungs from chest X-rays, which can be used in practice by clinical experts to diagnose and treat pneumonia. Model developed was trained, evaluated and tested using the Chest X-Ray Images (Pneumonia) dataset accessible on Kaggle repository. The main model comprises of 5 convolutional layers. The model attained accuracy of 92.78%, precision at 91.36%, recall at 97.69% and F1-Score at 94.43%. Dropout regularization is utilized in the model to limit overfitting in the completely associated layers. In nut shell, 5-layer CNN model developed from scratch instead of using transfer learning in this research can assist the clinical experts in evaluating the X rays for pneumonia detection with greater accuracy and taking up the treatment procedures at the earliest to avoid further complications and spread of disease.
Keywords - Convolutional Neural Networks, Pneumonia, Max-pooling