It has now become unsafe to go to the hospital every time we feel unwell, since there is a risk of getting affected by COVID-19. Also if the patient recognizes his/her symptoms, and if somehow we can tell him what is the disease he is likely to be affected with then he/she can take precautions accordingly. We believe that everyone should have easy access to great health care. Thus, there is a need to connect patients virtually with doctors.
What it does
Our team build a web-application that will help the people to know about the disease they are likely to be infected with based on their symptoms, then they can take the precautions accordingly. In case of high risk they can directly consult a doctor in an effective manner using the video calling feature of the application. Our application consist of 3 features: First, we designed a computer-aided diagnosis system(or disease prediction system) where users can get to know whether they are infected with a particular disease or not using machine/deep learning. For this, they are required to enter their medical details on the form or upload X-Ray/MRI image. Secondly, there is an option to enter the symptoms (simply type the symptoms or record the audio in browser) they are experiencing and the patients will get to know what possible diseases they might have along with the precautions that they must take. Third feature is the doctor appointment system wherein patients can not only search doctors based on region or specialization, but also connect virtually with the doctors around the globe.
How we built it
We used scikit-learn for training machine learning models while tensorflow/pytorch for deep learning and segmentation models. Dataset was obtained from kaggle website. A model was trained to predict the disease based on the user's symptoms. Sqlite database was used to store data of doctors around the world and the user can search doctors based on location or specialization, request for a video meeting and then discuss his/her problems.
Challenges we ran into
Connect patients with doctors in real-time. Improving the accuracy of ML/DL models.
Accomplishments that we're proud of
Although our application is not fully functional, we have designed a decent prototype which very well meets the requirements of our plan of actions i.e, fast redressal of grievances of patients from doctors in real time. Secondly, All the models trained have a accuracy of more than 95%.