Inspiration
Receiving clinical lab results—like a Complete Blood Count (CBC) or Comprehensive Metabolic Panel (CMP)—often leaves patients staring at a wall of dense medical acronyms, raw numbers, and unfamiliar reference ranges. That confusion frequently leads to unnecessary anxiety or uninformed internet searches before a doctor's appointment. VitaScan AI was inspired by the goal of bridging that gap in patient health literacy by turning raw clinical data into clear, empathetic, and actionable insights.
What it does
VitaScan AI is an intelligent clinical lab report interpreter that transforms unstructured blood test and metabolic panel text into an interactive, easy-to-understand health dashboard:
- Instant Biomarker Parsing: Users can paste raw lab report text—or test the app immediately with a one-click Load Sample Panel button.
- Overall Wellness Index: Evaluates the entire panel against clinical reference ranges to generate an at-a-glance wellness summary.
- Interactive Data Visualization: Renders clean, color-coded visual reference charts powered by Chart.js so patients can immediately see which biomarkers are optimal, borderline, or out of range.
- Plain-English Explanations: Breaks down what each individual biomarker measures and why it matters in everyday language.
- Doctor Consultation Prep: Generates tailored, specific questions for patients to discuss with their physician alongside practical next steps.
How we built it
- Backend: Built with Python and Flask to serve the web application and handle API routing cleanly.
- AI Engine: Integrated the Google Gemini API using structured prompt engineering to reliably extract biomarker names, numerical values, units, and clinical reference ranges from messy, unstructured text while returning consistent JSON payloads.
- Frontend & Visualization: Designed a responsive, clean clinical interface using HTML5, CSS3, and vanilla JavaScript, paired with Chart.js for dynamic, interactive biomarker range bars.
Challenges we ran into
- Handling Unstructured Text Formats: Lab reports from different clinics vary wildly in formatting, spacing, and unit abbreviations. Designing prompts that consistently normalize these inputs into structured JSON without hallucinating missing values required iterative testing and strict schema instructions.
- Balancing Medical Accuracy with Simplicity: Ensuring explanations remain clear and reassuring for everyday patients while accurately reflecting clinical reference ranges—and making sure the app clearly frames insights as educational support rather than a medical diagnosis.
Accomplishments that we're proud of
- Delivering a seamless, end-to-end full-stack AI prototype that processes complex multi-marker lab panels in seconds.
- Combining generative AI parsing with interactive Chart.js visualizations and tailored doctor questions in a single, intuitive interface.
- Creating a zero-friction user experience with built-in sample clinical data for immediate demonstration.
What we learned
- How to enforce strict structured JSON outputs from the Google Gemini API so frontend charts and UI cards render reliably every time.
- The importance of thoughtful UX design in healthcare applications—using clear visual hierarchy, plain language, and actionable takeaways to reduce patient overwhelm.
What's next for VitaScan AI
- PDF & Image OCR Upload: Adding direct support for uploading PDF lab reports and smartphone photos of printed results.
- Historical Trend Tracking: Allowing users to compare multiple lab panels over time to visualize biomarker improvements.
- Multi-Language Translation: Expanding plain-English explanations into multiple languages to make lab results accessible to diverse patient communities worldwide.