Inspiration
The idea for FeverSense emerged from the growing need to make early fever diagnosis faster, more accurate, and accessible. During seasonal outbreaks and post-pandemic conditions, healthcare systems often face high patient loads and delayed triage.
We wanted to create an AI-based system that assists clinicians and patients by predicting fever severity, recommending medication, and maintaining digital patient records. Our goal was to combine data-driven insights with clinical logic to improve decision-making efficiency.
What It Does
FeverSense is an AI-powered diagnostic platform that predicts the severity of a fever — whether it is Normal, Mild, or High — based on patient symptoms and vital parameters.
It then recommends appropriate medication and securely stores each patient’s record in a structured database for future reference.
The system uses a Random Forest model that takes a feature vector of clinical data such as temperature, heart rate, and blood pressure, and predicts the corresponding triage level.
Mathematically, for a given patient feature vector ( \vec{x} ):
[ \hat{y} = f(\vec{x}) = \arg\max \big(P(\text{Normal}), P(\text{Mild}), P(\text{High})\big) ]
where ( \hat{y} ) is the predicted fever severity level.
How We Built It
The project is structured into three main layers:
Machine Learning Model:
- The dataset was cleaned and preprocessed using
pandasandscikit-learn. - A Random Forest classifier was trained to predict fever severity, and a second model was trained to recommend medication.
- Feature scaling, label encoding, and data balancing techniques were applied for optimal performance.
- The dataset was cleaned and preprocessed using
Backend (Flask API):
- Implemented using Flask to serve ML predictions through RESTful endpoints.
- Handles routes such as
/predict,/history, and/delete_history. - Integrated with SQLite for patient data management and ensured error-free input handling for mixed data types.
- Implemented using Flask to serve ML predictions through RESTful endpoints.
Frontend (Streamlit):
- Developed using Streamlit to provide a clean and interactive user interface.
- Supports temperature input in both Celsius and Fahrenheit, displays predictions with confidence levels, and visualizes feature importance.
- Includes patient history viewing, record deletion with confirmation, and CSV download functionality.
- Developed using Streamlit to provide a clean and interactive user interface.
Challenges We Faced
- Managing categorical and numerical features efficiently during preprocessing.
- Ensuring seamless communication between Flask and Streamlit without CORS or latency issues.
- Preventing database locks during concurrent insertions and deletions.
- Maintaining interpretability while improving model accuracy.
- Designing a professional and medically intuitive interface suitable for hackathon presentation and real-world applications.
Accomplishments
- Successfully built an end-to-end AI diagnostic system capable of real-time fever triage and medication recommendations.
- Integrated data storage and retrieval with a reliable backend using SQLite.
- Created a functional and aesthetically appealing UI that supports secure data handling and visualization.
- Ensured clinical consistency in recommendations and implemented a safety override for medication logic.
- Delivered a fully deployable system that bridges AI, healthcare, and usability.
What We Learned
- Designing and deploying a complete healthcare AI workflow from data preprocessing to live deployment.
- Implementing robust data validation, categorical encoding, and error handling for medical datasets.
- Best practices for connecting machine learning models to real-time web applications using Flask and Streamlit.
- Understanding the role of user experience in health-focused technology solutions.
- The importance of explainability and transparency in AI-driven healthcare systems.
What’s Next for FeverSense
- Extend the model to detect and classify multiple infectious conditions beyond fever.
- Integrate Explainable AI (XAI) techniques like SHAP or LIME for improved model interpretability.
- Deploy FeverSense on cloud platforms such as AWS or GCP for real-time telemedicine support.
- Develop a mobile version to enable remote patient monitoring and quick triage assistance.
- Collaborate with healthcare professionals to refine datasets and validate predictions in real clinical environments.
Log in or sign up for Devpost to join the conversation.