Parkiscan

ParkiScan — An Educational AI Demo on Facial Motion Analysis EDUCATIONAL DEMONSTRATION ONLY ParkiScan is a student research project, NOT a medical device, NOT FDA-approved, and NOT a diagnostic tool. It cannot detect, diagnose, treat, cure, or prevent Parkinson's disease or any other condition. The output is a model demonstration only — not a health assessment, screening, or medical result. Always consult a licensed healthcare professional for any medical concerns. WHAT THIS APP IS ParkiScan is a hands-on educational demo that shows how artificial intelligence and computer vision can analyze patterns of facial motion. It was built as a student research project to help learners, educators, and curious users explore AI in a transparent, accessible way. HOW THE DEMO WORKS 1. Open the app and acknowledge the educational-use disclaimer 2. Tap "Start Demo Scan" 3. Record a short video of your face making natural expressions 4. The app extracts 468 facial landmarks and runs them through a machine learning model 5. View the model's raw output and feature breakdown for educational discussion WHAT YOU'LL LEARN ParkiScan is designed as a teaching tool that demonstrates: • Real-time facial landmark detection using Google's MediaPipe Face Mesh • Feature engineering from raw landmarks: - Eye Aspect Ratio (EAR) - Mouth Aspect Ratio (MAR) - Jaw angles - Movement velocity and acceleration • How a machine learning model produces an output from these features • Why ML model outputs include uncertainty and require careful interpretation • Why responsible AI in healthcare requires transparency, disclaimers, and human oversight EDUCATIONAL FEATURES First-launch acknowledgment that clearly explains the app's purpose Demo output labeled as model output, not a health result Educational breakdown of facial motion features "Known Limitations" section explaining where the model can fail Designed for classroom demos, science fairs, and self-study Encourages critical thinking about AI in medicine WHO THIS APP IS FOR • Students learning about AI, machine learning, or computer vision • Educators teaching applied AI or biomedical engineering concepts • Science fair and research project participants • Anyone curious about how AI models analyze visual data KNOWN LIMITATIONS (TRANSPARENCY MATTERS) • Trained on a small, non-clinical dataset • May not generalize across age, ethnicity, or gender • Sensitive to lighting, camera quality, and camera angle • Produces false positives and false negatives • Has not been validated against any clinical standard • Cannot detect, diagnose, or rule out any disease PRIVACY ParkiScan respects your privacy. Videos are sent to a processing server only to run the AI demo and are not retained. Saved demo outputs live only on your device, not in the cloud. IMPORTANT DISCLAIMERS • ParkiScan is NOT FDA-approved. • ParkiScan is NOT a clinical or diagnostic tool. • Demo outputs have NO medical validity and should NEVER be used as a health assessment. • The model can produce false positives and false negatives. • Do not use this app to self-diagnose or to evaluate anyone else. • If you have health concerns, consult a qualified healthcare professional — only a licensed neurologist can diagnose Parkinson's disease. BUILT WITH ParkiScan was built using open-source tools including Google MediaPipe, scikit-learn, OpenCV, Flutter, and Python. It is shared as an academic learning project, not a commercial product. OUR GOAL We hope ParkiScan inspires students and educators to explore how AI can responsibly support human health — while always recognizing that real medical decisions belong with qualified medical professionals. Thank you for using ParkiScan as a learning tool.