This project was completed for the Deep Learning Week '25 hackathon. For more information, please view the full README here!
Healthcare is shifting from reactive treatment to proactive, AI-driven prevention. Participants are challenged to develop AI-powered solutions that for example, integrate remote health monitoring, mental well-being support, or lifestyle management. Your solution can leverage wearable or mobile technology to track real-time health data, detect patterns, and provide intelligent alerts for early intervention. It may also incorporate AI-driven mental health support, offering personalized insights, virtual assistance, or community-driven engagement. Your system could promote healthier lifestyle choices by analyzing behaviors and recommending tailored interventions to prevent chronic diseases.
Build an AI-centric, seamless, and adaptive solution that empowers individuals and transforms healthcare accessibility.
Nowadays, wearable technology has become increasingly accessible and people are having easier and cheaper access to their body vitals and metrics, through devices such as smartwatches and fitness trackers. We can leverage this increased accessibility of IoT wearable devices, and use the easily accessible data to perform data analysis for users’ health.
Currently, users do not have access to early detection and warnings as well as preventive measures for diseases. Usually, when patients find out about their illnesses, it is already too late. Frequent health screenings can solve this issue, but they are expensive and time-consuming, and are usually only done on a yearly basis.
Leveraging on the myriad of data that smartwatches provide to users, we aim to create an iOS app with an accompanying watchOS app that provides data analysis for early disease prediction. The app will display early warnings to users and preventive measures they can take to combat these illnesses. Deep learning algorithms will be used to analyse past patient data and create a model that associates anomalous patient vitals with early disease symptoms. This model will then be run on real-time user data, analysing their historical body vitals and comparing it with their current body vitals, to determine whether there are anomalies corresponding to specific diseases.
| Dataset | Regression | Random Forest | Deep Learning (ANN) |
|---|---|---|---|
| Diabetes | 91.6% (Order 2 Ridge) 92.3% (Logistics) |
90.4% - 91.8% | 92.7% |
| Stroke | 97.7% | 95.1% | 100% |
| Heart Disease | 83% | 85% | 50% |
| Sleep Disorder | 86% | 87% | 85% |