Pneumonia Detection via Chest X-Rays

Abstract

Although pneumonia is a treatable disease, it kills more children than any other disease or infection. According to a UNICEF report, Pneumonia claimed lives of 808,694 children under the age of five in 2017[1]. In Pakistan, an estimated of 58,000 children died of Pneumonia in 2018 placing Pakistan at 3rd number in most deaths by pneumonia in the world [1]. The major cause of deaths by pneumonia in Pakistan is low availability of doctors and medical personnel in certain cities and areas. So, we are developing a web-based application that can detect pneumonia from the image of a patient’s chest (lungs) x-ray. With unavailability of doctors in certain regions, patients need to travel long distances to get doctor appointments wasting crucial time. This project is aimed to assist patients to get diagnosis fast from home and act relatively. We are using deep learning algorithm called Convolutional Neural Network (CNN) that will be trained and tested on a dataset of chest X-rays. Users would simply upload their chest x-ray image on the website and the CNN algorithm will predict if the patient has pneumonia or not from the x-ray image. With even low availability of doctors in this covid pandemic this project can really help patients to get diagnosed in time as patient’s condition keeps getting critical with time. Not all cases of pneumonia are serious, in some cases the patient has mild symptoms so in that case they should not panic but in case of moderate and severe pneumonia, the patients need to seek immediate medical assistance so that treatment can be started in time. Therefore, the deep learning model will then predict its severity as “Mild”, “Moderate” or “Severe”. The website will use the users’ current location from their GPS and use that location to find pulmonologist near their current location and display their details with location of the pulmonologist using Google Maps API so that the user may make appointments for further diagnosis and treatment. Let us hope the project helps patients as we think it would.