Corona Virus (Covid-19) is a Global Pandemic which has killed ~2+ Million people and has put the lives of Billions around the world at risk. So, it’s highly important to have a cost free easy to access diagnosis solution to detect this disease from the comfort of your Mobile App (Cell Phone) , as its contagious to go out and get tested.
The solution should also be affordable by all.
Early detection can lead to easy protection.
Hence, we have come up with a solution to diagnose Covid-19 symptoms and check your risk from Mobile App.
What it does
This is a Diagnosis Mobile App for Covid-19, it can be used to detect the risk.
Our Artificial Intelligence (AI) Cough based Mobile App solution can help in early diagnosis i.e. when the patient observes she/he has symptoms of Covid-19 he can use this easy to use Mobile App for diagnosis and see if he is at risk of Covid-19. In this Mobile App he can record his temperature, dry cough, shortness of breath and other symptoms and hit submit. Once he submits his inputs, the App takes these input symptoms and returns the output Diagnosis report. i.e. if he is at risk of Covid-19, so that he can take timely next measures to protect.
The App takes the input symptoms and cough sound provided by the user and sends them to AI trained Machine Learning (ML) classification model and the decision program in the App. The cough patterns are then matched with those of corona virus patients in the machine learning model and this result is combined with other symptoms in the App program which returns the risk of the user. The App returns the diagnosis report to show the users corona virus risk.
How we built it
This App is an Android based Mobile App. Which uses Java and Artificial Intelligence (AI) - Machine Learning (ML) Python Classification Model to determine the risk of patient for Covid-19.
Mobile App’s front-end is built in Android design Studio with XML. Java is used for developing back end which has the core functionality. Cough sound processing is done with a Tensor Flow (https://www.tensorflow.org/) trained AI - ML Model. This Model is written in Python. The model classifies the cough i.e. if it matches or does not match with Covid-19 cough.
The Python ML code created is hosted in Amazon AWS EC2 Cloud Server. The Android application and Python model are connected using REST APIs. Android application will pass cough sound to python model in the cloud. Then the model classifies the cough i.e. if it matches or not with Covid-19 cough. This result is then combined with other symptoms in Java code which gives the output if the patient is at risk or not.
*The AI-ML Python and App code are included in the Code/Attachments in GitHub link
Software: Python, Tensor Flow, Amazon AWS EC2 Cloud, Android 10 Studio, Java, REST API, XML
Hardware: Android Cellphone
Concepts: Artificial Intelligence (AI), Machine Learning (ML), Object Oriented
Other Inputs: Cough sound, Thermometer
Input: Symptoms -fever, cough, shortness of breath, fatigue, body pains, headache, loss of smell, sore throat, congestion, nausea, diarrhea
Output: Diagnosis and Cough report
Challenges we ran into
Some of the challenges we ran into were, while connecting Android application to AI model. For which we tried different methods including on premises, drivers, and then newly tried by hosting Model on AWS EC2 Cloud Server. Which finally worked, and we were able to connect the application with model.
Accomplishments that we're proud of
We were able to overcome our initial challenges while building the model and connecting to Android App We were finally able to build our Product per our vision and strategy and make this Diagnosis Mobile App.
What we learned
Perseverance for right vision pays
What's next for Mobile App for Corona Virus Diagnosis
· Use real time Covid-19 patients cough data for ML model
· Introduce noise cancellation to improve cough feed quality
· Conduct Clinical trials on Covid-19 patients with the App
· Include respiration rate and oxygen saturation level measurements in the App
· Make it as approved Covid-19 Diagnosis App available for all in Android and iOS
and other online research material