PupilMetrics Research

PupilMetrics Research is an educational application designed to help users explore and learn about pupil characteristics through interactive image analysis with machine learning (ML) integration. The app offers tools to capture, observe, compare, and visualize pupil features from videos and photos, presenting metrics solely for learning, research, and educational purposes. Core PLR Recording & Analysis Features Dark-Adapted Baseline Capture - Automatic 3-second dark baseline before each flash for maximum dynamic range and accurate constriction measurement. Single-Trial PLR Recording - Structured protocol with 200 ms calibrated bright flash + 2.6-second recovery window using Dino-Lite iriscope. Automated Pupil Detection & Tracking - Hybrid classical CV + ONNX ML ensemble for robust pupil/iris ratio measurement across all frames. 5-Point Median Smoothing - Noise reduction to eliminate blink and detection artifacts. PLR Grade Assignment (A–F) - Automatic grading based on signal-to-noise ratio and constriction amplitude with clinical interpretation. Key PLR Metrics Calculation - Baseline diameter, constriction amplitude (%), latency, constriction velocity, redilatation velocity, PIPR (Post-Illumination Pupil Response). 3-Trial Habituation Protocol - Consecutive trials with enforced rest intervals to measure cortical modulation. Cross-Trial Waveform Overlay - Visual comparison of three PLR curves on a single timeline. Bilateral Scanning & Anisocoria Detection - Compares right (OD) and left (OS) eyes with severity grading (Physiological / Mild / Significant). Monitoring & Clinical Workflow Features Serial Trend Monitoring - Longitudinal comparison of PLR metrics across multiple sessions/visits. Session History & Database - Local SQLite storage of all scans with search and reload capability. Local Processing Only - No cloud connectivity required; all analysis runs on-device with full patient data privacy. Dino-Lite Hardware Integration - Direct control of LED, frame capture, and calibrated flash via 32-bit C# COM bridge. Fits an ellipse to the dark region using covariance matrix eigenvalue decomposition Extracts boundary points by ray-casting from center outward Returns center, major/minor axes, orientation angle, and boundary points Analyzes 72 boundary points (every 5 degrees) around the pupil edge Calculates deviation from the average radius at each clock position Groups deviations by clock hour (12 zones) Identifies flattenings (inward deviations) and protrusions (outward deviations) Strong Ethical & Legal Safeguards - Built-in disclaimers, no specific drug identification, emphasis on clinical context only. Important Disclaimer: This app is strictly a research and educational tool. It is not a medical device, does not provide medical advice, diagnosis, clinical interpretation, or evaluation of health conditions, and is not validated for clinical use. Results are informational observations only and should never replace professional healthcare consultation. Always consult qualified healthcare professionals for any health-related concerns. Intended Use: Educational demonstrations Research exploration Learning visual pattern recognition Academic or training purposes To review PupilMetrics App full documentation, please visit: https://cnri.edu/