Inspiration
Early detection of melanoma drastically improves 5-year survival rates, yet single-snapshot AI diagnostics often create unwarranted panic or false reassurance. Clinical dermatologists rely heavily on the "E" in ABCDE: Evolution—tracking how a lesion changes over time. We built DermaTrack to give patients and clinicians a free, accessible, and mathematically grounded way to track millimeter-accurate skin lesion progression using everyday hardware.
What it does
DermaTrack is a progressive web platform that standardizes skin imaging and analyzes longitudinal lesion drift:
- Reference Metric Calibration: Overcomes smartphone focal-length distortion by normalizing pixel dimensions against everyday circular references (e.g., standard coins or 20mm stickers) to compute true Pixels-Per-Millimeter (PPM).
- Automated ABCDE Biometrics: Leverages native OpenCV algorithms for Otsu lesion contour segmentation, dual-axis centroid symmetry folding (A), isoperimetric compactness $P^2 / 4\pi A$ (B), multi-spectrum pigment variegation (C), and calibrated millimetric diameter (D).
- Longitudinal Evolution Engine: Centroid-registers sequential visits ($T_1 \rightarrow T_2$), computing growth velocity ($\Delta \text{Diameter}$, $\Delta \text{Area \%}$) and rendering a differential hot-crimson peripheral expansion heatmap.
- Doctor Consultation Intake: Generates a structured 1-page clinical summary PDF complete with targeted questions for the dermatologist.
How we built it
- Frontend: Next.js 16 App Router, React 19, Tailwind CSS, Lucide icons, and zero-shift CSS clip-path comparison sliders.
- CV Microservice: FastAPI backend running native OpenCV (
cv2) and NumPy for Hough circle transforms, bilateral edge filtering, spatial moments, and morphological matrix operations.
Challenges & Accomplishments
Standardizing photos taken under different camera angles without dermoscopic hardware was our greatest technical hurdle. By implementing automated Hough Circle detection coupled with an interactive millimeter calibration ring, we achieved reliable metric scaling without specialized clinical equipment.
Built With
- cancer
- computer-vision
- detection
- digital-health
- fastapi
- next.js-16
- numpy
- oncology
- opencv
- python
- react-19
- tailwind-css
- typescript
Log in or sign up for Devpost to join the conversation.