The real time reverse transcription-polymerase chain reaction (RT-PCR) detection of viral RNA from sputum or nasopharyngeal swab has a relatively low positive rate in the early stage to determine COVID-19. Due to this, an infected person might test negative in the earlier stages even if he/she is carrying the infection in their system. The manifestations of X-Ray imaging and computed tomography (CT) of COVID-19 have their own characteristics according to recent research and observations published online, which are different from other types of viral pneumonia, such as Influenza-A viral pneumonia. Therefore, the need of the hour calls for an early diagnostic criteria for this new type of infection as soon as possible. This web application is aimed to establish an early screening model to distinguish COVID-19 cases from other Influenza-A viral pneumonia and healthy cases with X-Ray and CT images using deep learning techniques. A person who gets tested positive for the disease using our system can take an immediate initiative to self-quarantine himself/herself and avoid the infection from spreading to other people since the virology tests can take a significant amount of time to give concrete results, especially due to the hospitals and health facilities of USA being overwhelmed with a large number of COVID-19 cases.

What it does

It is simple. Just upload a JPG/PNG image format of your X-Ray or CT Scan and click Submit! And Walaa! Our model tells you whether you have tested positive or negative for COVID-19 along with the prediction score associated with it.

How I built it

The Machine Learning model was trained using auto-training tools such as Teachable Machine by Google using training data sets collected from databases of several health facilities such as: 1) Multiple open source GitHub repositories

A web-application was created using HTML, CSS, JQuery, Python and hosted on a Flask server hosted on Heroku to interface with the user and interact with the model for POST requests.

Challenges I ran into

The main challenge was to build an accurate model with the scarcity of training data sets. Since the virus is very new, online databases and sources have a very scarce amount of X-ray and CT scan samples of COVID-19 patients.

Accomplishments that I'm proud of

The main accomplishments which I am proud of is that the model has an accuracy of nearly 87.6% in correctly detecting COVID-19 infected scans, which can be further improved upon by proper and accurate segmentation of training datasets.

What I learned

Learning how to use technology and deep learning to tackle critical global problems which can help save lives.

What's next for COVID-19 Detection Using Deep Learning

Will be working on training our Deep Learning model with other image formats such as DCM, DICOM, TFF which are more heavily used in medical facilities for taking X-rays and CT scans.

Try It out



css, flask, html, jquery, numpy, python, tensorflow

Devpost Software Identifier