Inspiration: We’re a team of all girls who find the intersection between healthcare and technology fascinating, and we wanted to work towards solving an issue that primarily women face. The Hacklytics Team provided us a Kaggle Breast Cancer dataset, and we decided that this perfectly fit our goals.
What it does: Intakes medical records and blood test samples and tests the given information with our trained model to determine which type of breast cancer someone may have, as well as an AI-driven treatment and resource provider.
Challenges we ran into: A significant challenge we ran into was finding sufficient data for the ML model. To expound, we noticed that the ratio of those with Infiltrating Ductal Carcinoma to those with Infiltrating Lobular Carcinoma AND Mucinous Carcinoma in our data set was an extreme difference. This affected our model accuracy at one point because it became very good at predicting Infiltrating Ductal Carcinoma but not the others. We also had trouble integrating OpenAI with the model.
Accomplishments that we're proud of: We’re proud to have achieved a functional yet attractive UI, a neural network model with 90%+ precision, and integrating OpenAI to provide further insights.
What we learned: We learned how the importance of having a strong dataset as a foundation and double checking that all dependencies are installed for the type of OS.
What's next for Breastie: If we were to have more time, we would have liked to utilize mammograms or a similar descriptive medical image of the breasts in order to elevate Breastie into a multimodal agent for deciphering between the different types of breast cancer. We would also like to develop Breastie into a mobile app in addition to the current web app.
Log in or sign up for Devpost to join the conversation.