A Pandemic is commonly described as a disease, which can rapidly spread across regions - not limited to infecting cities, provinces, countries, and possibly continents. Currently, the world is facing the COVID-19 Pandemic caused by a novel Coronavirus that originated in Wuhan, China. As of May 3rd, there have been a total of 3.45 million confirmed cases and 244,000 deaths. In cases of highly infectious pandemics, it is crucial to implement early detection and immediate communication between doctors and patients for public health and safety. This is why we have created COV-Vision.
What it does
COV-Vision is a web application that allows doctors to diagnose diseases and send diagnostic results automatically to their respective patients. A user first logs-in with a Google Account. If they are a doctor, they would have the ability to upload their patient's x-ray images onto our platform. This image runs through a convolutional neural network (CNN) in the node.js servers which classify if the patient has COVID-19 or not. Then, the doctor is able to send the results directly to their respective patients via our platform. If you the user is a patient, they will automatically receive the diagnostic results from their doctor with all the details alongside the x-ray image. This system will remove the need for patients to go back to the doctor's office to receive their x-ray results. If the patient is COVID-19 posititve, safety tips are given on precautionary measures to avoid spreading the virus as well as helpline numbers in case of difficulty. If the patient does not have COVID-19, tips on social distancing and maintaining proper hygiene are given. Furthermore, this system can be rapidly implemented in the cases of novel pandemics as well as act as an early detection and warning system for disease conditions. COV-Vision allows for a fast and effective way for diagnosing disease conditions and allowing communication with doctors and patients in a seamless environment.
How we built it
The front-end is built using Svelte and our back-end is hosted on a node.js server as well as using Firebase Firestore to store user status. We used a convolutional neural network (CNN) made using TensorFlow which is transferred over to the node.js server.
Challenges we ran into
The biggest challenges we ran into were in the machine learning and back-end components of our project. The machine learning component had a very low accuracy at first. However, we were able to quickly fix this problem by adding more datasets and layers to the CNN. The back-end component of linking everything together was hard. Eventually, after looking over a lot of documentation - we successfully linked the front-end with the node.js servers and the machine learning component.
Accomplishments that we're proud of
We are most proud of having accomplished many of our goals for this application. We were successfully able to integrate all of our back-end and front-end components together despite the struggles. The application itself has many functionalities for the users. Overall, we have put our knowledge and experience with these technologies and languages to excellent use.
What we learned
We learned a lot about optimizing CNNs for the purpose of image recognition. We are thankful by the helpful tips which were provided to us from our mentor (shout out to my boy Shun Lin!). In terms of the back-end, we had to learn a lot behind connecting a CNN with a node.js server to analyze the images and send them to respective users. This project helped us learn a lot about advanced back-end networks and utilizing them in an efficient manner.
What's next for COV-Vision
In the future we would like to add more functionality rather than just detecting COVID-19 from lung x - rays. We would like to add more detailed analysis of various other disease conditions such as pneumonia, tuberculosis, and malignant tumors from a wide variety of patient data.