NexaHealth

About the Project — NexaHealth

Inspiration

Our team was inspired by how difficult it can be for people to understand the severity of a wound, especially when they do not have quick access to medical help. We wanted to create a tool that gives users an instant assessment, simple guidance, and a clear path to care. We were also interested in using blockchain technology to help people control their own medical data safely.

What We Learned

Throughout this project, we learned how AI can analyze images, how wound-classification models work, and how to design a basic medical triage system. We also practiced integrating different technologies, such as:

• Image analysis with machine learning
• A Telegram chatbot for fast communication
• Blockchain-based identity for secure data use
• Location-based search for nearby medical centers

I also learned how important user experience is, especially when dealing with medical information.

How We Built It

We broke the project into several steps:

1. Wound Classification Model We used a wound image dataset to train an AI model to classify different wound types. We used basic preprocessing and a simple CNN architecture. Example math expression: We minimized the loss [ \mathcal{L} = -\sum_{i=1}^{N} y_i \log(\hat{y}_i), ] where ( y_i ) is the true class and ( \hat{y}_i ) is the predicted probability.

2. Severity Assessment System Users upload an image, rate pain, and select symptoms. We combined these inputs with the model’s prediction to estimate wound severity.

3. Triage and Location Guidance Based on severity, the system recommends self-care, urgent care, or emergency care. We used a simple API call to find nearby medical facilities.

4. Blockchain-Based Data Identity We experimented with generating a unique user ID from hashed medical data so users can access the system without traditional login credentials.

5. Telegram Chatbot We connected the core features to a Telegram bot so users can quickly get translations, assessments, and directions.

Challenges We Faced

We faced several challenges during development:

• Some wound images in the dataset had low resolution, which made classification harder.
• Integrating all components (AI model, chatbot, location search) within a short time was difficult.
• Designing a simple triage system without giving medical advice was a careful balance.
• Implementing blockchain identity in a short timeframe required us to simplify the design.

Despite these challenges, building NexaHealth taught us how to combine AI, user interaction, and secure data handling into one meaningful tool.

Built With

• flask-cors>=4.0.0
• flask>=2.3.0
• google-generativeai>=0.3.0
• joblib>=1.3.0
• matplotlib>=3.7.0
• numpy>=1.24.0
• pillow>=10.0.0
• pytelegrambotapi>=4.0.0
• python-dotenv>=1.0.0
• requests>=2.31.0
• scikit-learn>=1.3.0
• seaborn>=0.12.0
• tensorflow>=2.13.0
• werkzeug>=2.3.0