As a son/daughter living away from parents, one would always wish that there is someone looking after their elderly parents. That, and coming across an article about the millions spent by insurance companies to verify insurance claims made me realize that there's no single suite that provides a complete solution to healthcare needs.
What it does
MediX collects basic metrics such as Heart Rate and Blood Pressure from Smart bands and creates a Model about a person's ideal vitals.
It then uses a preliminary Anomaly detection algorithm to detect anomalies in body vitals, alerting a doctor who can have a look at them and prescribe medicines online if a visit isn't required. Thus having someone always keep an eye on the health of your loved ones and enabling early detection of various ailments.
MediX also provides an Alexa Skill which reminds users about the medicines they need to take.
MediX also provides a Chatbot for users to describe their ailments and get connected to a doctor.
MediX stores all the prescriptions generated on a blockchain, thus enabling easy verification of the authenticity of medi claims saving millions. Not only that, but it also helps pharmacies verify authentic prescriptions thus preventing the unauthorized sale of drugs via forged prescriptions.
How we built it
We used flask as a backend and for rendering pages server-side rendering with MongoDB acting as a database. We also used SciKi-learn for training the model. We hosted the Alexa skill on AWS Lambda written in python using Alexa skills SDK Figma for prototyping and zurb-foundation UI Kit.
Challenges we ran into
We couldn't find a dataset with required body vitals from fitness bands, so we took help from a fellow hacker on discord who agreed to share data from his Fitbit
Most modern blockchains require multiple nodes, which won't be feasible for a hackathon project. so we had to develop our own rudimentary implementation of blockchain.
Spent a lot of time designing a unique UI and experience for the users.
Hours wasted on outdated Alexa Documentation and Myriad of outdates Google Oauth tutorials
Accomplishments that we're proud of
Implementing our own simple blockchain. Getting about 13% False Positive Rate on our model.
Making a lot of new friends ( Kind of a big thing for us introverts )
What we learned
The official documentation is better than a random tutorial ( for like the 10th time ).
Set Priorities for tasks and do them accordingly
Working together in a team with different ideas is a huge challenge in itself not to mention the all-nighters pulled off.
What's next for MediHelper
Using data aggregated to analyze and track epidemics using patient symptoms, reports and other metadata such as location, age, etc.
Collaborating with Major Pharmacies and Insurance firms to make the whole process more seamless