Title
: COVID-19 Detector from Chest X-Ray Image Analysis
Description
: The recent outbreak of coronavirus (COVID-19) has encouraged us come up with the solution “eGeneration BeatCorona COVID-19 Screening Tool” (http://beatcorona.egeneration.co) to help people of Bangladesh. This solution can be used by the local people from their home in the time of COVID-19 pandemic. Users can effortlessly perform COVID-19 screening from anywhere online providing their chest X-ray image. This will be useful for the people, as our country have very limited amount of testing kits available. Our solution has already been featured in government’s corona related website (https://corona.gov.bd/screening-tools).
Let’s discuss what we have in our solution:
- AI based COVID-19 screening using chest X-ray image.
- Real-time COVID-19 statistics for Bangladesh and worldwide.
- Visual representation of the total number of affected, the dead and cured cases.
- Steps to prevent coronavirus from spreading.
- Awareness about the symptoms and precautionary steps to take.
The main features in detail of the solution:
- Using the AI based screening tool, users will be able to identify if they are healthy or have normal pneumonia or COVID-19 providing the X-ray image of the lungs.
- There are 4 categorized datasets by which the tool is trained: Normal, COVID-19, virus-pneumonia, bacteria-pneumonia. The training dataset consists 1000 of images of these four categories.
- The data are collected from various hospitals of Italy and China which are giving treatment to COVID-19 patients from the beginning of the pandemic. 4 categorized data helps to differentiate between the symptoms of healthy, normal pneumonia and COVID-19.
- Two machine learning models are used in this screening tool: Fit-forward Neural Network and Convolutional Neural Network. Among these two, Convolutional Neural Network seems to give promising accuracy on the testing dataset.
- We also collect information of patient’s age, sex, location etc. which can help us on further analysis.
How our solution can have impact in this situation?
- People can perform COVID-19 screening from home only using chest X-ray image
- Further guidelines have been provided according to test report
- Safety measures have been provided for general people
- People are updated with real-time COVID-19 statistics
Guideline to use the solution:
- People can access this solution online. All you need is a mobile phone/desktop and an internet connection.
- Users have to input their name, age, sex, location, contact number and an x-ray image of their lungs.
- After they have submitted all the required things, they will be given a test result of what type of symptoms they have.
- From this, the users can proceed for further treatment, if required.
Future developments and time plan:
- In our detection tool, we are using two neural network predictive models. Moreover, we will be integrating two more models: Recurrent Neural Network and Deep Learning to achieve better performance.
- We will train our models with 5000 image data instead of 1000 to get more accuracy in the result.
- Additionally, we are planning on providing our user data to the government for further envision and analysis.
Our experimental results suggest that our model can reliably identify COVID-19 cases with almost 90% accuracy. There are plans for future improvement of the model and dataset. As our solution is already online and constantly getting user data, within two weeks we are confident to update the solution to an improved version with better accuracy.
