Calyx

Inspiration

30% of chronic wounds lead to hospital readmission because warning signs are invisible at home. We built Calyx to serve as the critical middleman between recovering patients and medical teams, turning "photo galleries" into actionable clinical data.

What it does

Calyx uses Computer Vision to act as a "digital ruler," measuring wound area and tissue health from a smartphone. It cross-references these scans with a clinical knowledge graph to alert doctors if a patient's biomarkers—like high blood sugar—are stalling their recovery.

How we built it

The platform is powered by a Python backend using:

The Eyes: OpenCV and K-Means for RYB tissue segmentation.
The Brain: NetworkX to map clinical comorbidities like Diabetes and Hypertension.
The Interface: Streamlit for the provider and patient dashboards.

Challenges we ran into

Achieving calibration precision across different cameras was difficult. We solved this by using a calibration coin as a real-world reference to calculate area in cm2 regardless of camera distance or angle.
Clinical Knowledge Mapping: Translating static medical literature into a dynamic Directed Graph required complex edge-weighting. We had to ensure that the BFS traversal correctly prioritized life-critical factors without creating "alert fatigue" for doctors.

Accomplishments that we're proud of

Computer Vision Precision: We successfully implemented a specialized OpenCV pipeline that moves beyond simple photo storage.
By utilizing K-Means clustering and RYB segmentation, we converted raw pixels into quantitative medical data, allowing us to track healing with a level of precision that visual estimation simply cannot match.

What we learned

Knowledge Graphs: We learned exactly how Knowledge Graphs work by using NetworkX to map out complex relationships and navigating them with BFS traversal logic.
K-Means Clustering: We learned how K-Means Clustering works by using it to segment wound images into distinct color zones, allowing us to quantify different tissue types automatically.

What's next for Calyx

Real-Time Processing: We want to implement Live Webcam Capture for real-time image processing instead of simple file uploads. This will allow us to handle frame-by-frame data in OpenCV to provide instant feedback on calibration and quality.