Diagnosis of Pneumonia from chest-x ray Images

Authors: Dr. Amol S. Dange, Shubham S . Babar, Kiran B. Dombale, Abhishek R. Mohite, Pranav N. Patil

Abstract: Computer Science & Engineering ADCET, Ashta Ashta, India — Pneumonia is an acute respiratory illness that necessitates appropriate and timely diagnosis to assist management. This project, Diagnosis of Pneumonia from Chest X-Ray Images, adopts deep learning techniques to enhance and support automation of pneumonia detection and classification using imaging technology. The system was developed with a reasonably easy to curate dataset of registered chest X ray images to develop an advanced Convolutional Neural Network, a ResNet-50 based model for measuring features of the chest X-ray image and a additional sequential CNN for predicting pneumonia from the extracted features of an image. The end user was presented with a suitable graphical interface, developed in react, to upload chest X-ray images at personal convenience. The system implemented an automated deep learning pipeline visioned to classify pneumonia types, suggests personal recommendations for further medical needs. The AI metrics of this model was an additional phase of the project application initiating a conversational chatbot interface that enables more user engagement through API and additional accessibility for interaction with other modules. An innovative doctor recommendation module suggests clinical specialists’ users may access based on the severity of the condition presented to the user. The model was also continued to be subjected to continued testing and validation using a real-time collected dataset, measure and examine performance metrics for verification of effective and efficient performance. In general, this project is likely to improve the management of pneumonia screening by diagnosing pneumonia faster and more accurately, making it a accessible choice for management of early pneumonia identification, severity classification and patient outcome automation using AI designed health care.