Pneumonia and other lower respiratory tract infections are the leading cause of death worldwide. Exposure and inhalation of the contaminated air in these environments ultimately leads to inflammation, and fluid filling in the lungs, in turn, reducing oxygen flow to the bloodstream . Aspiration pneumonia, pneumonia acquired by patients in hospitals (via contact with ventilators, instruments) are other categories of pneumonia. Viruses are the primary cause of pneumonia in children under five years . Children, infants, elderly, people with weakened immune systems, and people with severe alcohol misuse have an increased risk. Additionally, MRI scans and imaging facilities are expensive and obtaining an accurate diagnosis is cumbersome. Patients in fast-growing metropolitan cities have better access to diagnostic imaging facilities, whereas those in third world countries or rural areas and low growth economic regions do not have easy access to diagnostic imaging facilities. I worked on the application because I empathised with the people who suffer a disease like this with zero resources to cure.
What it does
The iOS application is developed keeping in mind the solution irrespective of the platform. The application uses the Apple Developer provided API’s and training the CreateML with CoreML model. The Application has an option for the user to click or upload Chest X-RAY images while accessing the mobile gallery as well. The format of the image input is specified and the app is integrated with InceptionV3 to accept only specific format. The application then analyzes the input image and gives a result pop up notification. If the person is diagnosed positive, they are redirected to the screen with doctor and hospital references which are responded using the person’s geo-location. If they are not diagnosed with Pneumonia, they are redirected to screen with Symptoms, Precautions and Preventions.
How I built it
The X-ray image data is analysed with the help of the trained model in the CoreML Model To predict the probability of pneumonia in the test data presented by the user. Also, preliminary advice on dealing with the disease is provided to the user. In the scenario where there is an active internet connection, the user can get connected to a group of expert medical practitioners (situated at distant metropolitan regions or anywhere in the world) available online, who can provide a real-time opinion for accurate prediction of the disease and affirm or disagree with the results provided by the application. The system is developed in such a manner that there is always a pool of doctors online, to serve the users of this application. The design methodology used in the development of this application for the preliminary detection of pneumonia is based on the CreateML and CoreML Apple API’s.
Challenges I ran into
Accomplishments that I'm proud of
Recently, my research paper - APPLICATION FOR DIAGNOSIS OF DETECTION OF PNEUMONIA DISEASE FOR RURAL CONTEXT has been accepted by International Journal of Scientific and Engineering Research (IJSER) to be published on 12th December. I am extremely proud that my intention of providing a solution irrespective of what technology i used has now got some recognition.
What I learned
Over the course of this project, I have learnt how important it is to research for real life problems that technology can be used to solve. This was also my first attempt at integrating Machine Learning to an iOS application and also learning how to access gallery and camera on devices with specific format. I learnt that how empathyzing with users leads to genuine solutions.
What's next for Pneumonia Detection Application (iOS)
Currently the application has an accuracy of 94.2% but I wish to increase the accuracy rate. My goal and aim for this application is to make it available and approachable to Rural Places or third world countries, where the citizens lack in financial resources to get there selves tested via chest XRAY's.