As our generation grows and technology advances, our world is becoming increasingly more fast-paced. Aging is a concern that gets put on everyone’s back burner, but nevertheless affects our lives through our loved ones.
It would be great if we could be by our parents' and grandparents’ side at all times to monitor their health, but this becomes increasingly close to impossible as most people are tied down by school, work, or other obligations. As a result, the elderly could purposefully downplay the severity of their own health conditions out of consideration for their caregivers, or perhaps be oblivious to the implications of their symptoms altogether. This is a frequent cause for individuals to miss the critical time frame to get the appropriate medical help, leading to consequences ranging from minor inconveniences to tragedy for the whole family.
What it does
With HomeHealth, we would like to provide a way for our users to do a medical pre-assessment based on symptoms they are experiencing and their basic health information.
Our program would provide the user with an analysis of possible conditions that they might be experiencing, encouraging them to see a physician in a more timely manner if necessary.
How we built it
For getting our Machine Learning Model to predict whether an individual tends to have Cardiovascular Disease or not, We first decided to get a dataset for Cardiovascular Disease, and then we tried to refine the data and dropped some unnecessary features like that it does not account into prediction. Later on, we started to structure the data by splitting it into Train Set and Test Set and also making sure that there are no duplicate entries as it might lead to overfitting the model and also took into consideration that our training data and test data comes from the same distribution of dataset after shuffling the dataset. Later on, we tried to work with different learning algorithms and even tried out using Neural Network for the same.
After taking into account several metrics like F1 Score, recall, and precision, we decided to go with Support Vector Machines and successfully completed it and getting around 68% accuracy. Later we tried to look into things and enhanced our model by tweaking it a little bit after which we were able to achieve an accuracy score of 73%.
To create an intuitive design for each of our users, simple and clear interfaces was required. We decided on an implementation of a few buttons, as well as clear indication and information on the screen. After wireframing, we used Adobe XD to iterate and prototype the final design.
We used HTML and CSS to build the basic structure of the webapp, as well as JQuery to manage some simple functionalities.
Challenges we ran into
Some challenges that we faced include the inaccessibility of comprehensive datasets regarding patient symptoms and diagnosis. Our training model could perform a lot better if we had more resources and time to perform the analyses.
We also didn't have enough time and experience to actually build the connection between the data analysis that we conducted via ML and the frontend webapp that we built.
Accomplishments that we're proud of
We were able to successfully train and model the data with our machine learning algorithm and get about 73% accuracy after analyzing the performance of our model through algorithms like RandomForestClassifier, K Nearest Neighbors, and Support Vector Machines (SVM), out of which we chose SVM over the other models as we want our model to be precise on the patients' data and were able to achieve that part.
We are also proud of our design of the webapp, as it is a clean interface that we believe is easy to navigate, while still having some flair in the look!
What we learned
Time management was the biggest key lesson in this Hackathon, where brainstorming ideas and then building upon that within a very short span of time is itself a big challenge. Team coordination is another factor that we improved on throughout the experience, as almost all the team members were from the different parts of the world and to have coordinate with everyone being in different timezones was key to our project. It was also a great learning experience for us to have a team of people from different expertise to build a final project, where our members come from different domains such as Bio-Medical Engineering, Computer Science and Engineering, and User Experience/User Interface (UX/UI).
What's next for HomeHealth
HomeHealth is planning to expand to assess an increasing variety of illnesses. With the right resources, we will be able to create more options for symptoms and detailed questions. Our purpose is to create a platform for our elderly that is as close to their needs as possible.
In order to do that, we also need to conduct user tests to see if the design is understandable and comfortable for them and for their caregivers.