研究目的
To develop a simplified check system to screen for the presence of strabismus, apart from the type or amount of ocular deviation, by digitalizing the cover-uncover test and implementing an abnormality determination process.
研究成果
The developed system demonstrated higher detection capability for ocular misalignment compared to conventional manual cover tests, with an 18% inconsistency rate against Maddox rod tests. Future improvements should address blinking interference and extend detection to vertical movements and abnormality extent quantification. The technology has potential applications in personal authentication and eye tracking.
研究不足
The system cannot distinguish between liquid crystal shutter operations and blinking, leading to potential erroneous judgments. It is currently limited to detecting horizontal eye movements and may not detect very small abnormalities (e.g., 1-2 prism diopters) or certain types of strabismus not detectable by cover test principles. Environmental lighting conditions (663 lx was optimal) affect performance.
1：Experimental Design and Method Selection:
The study automated the cover test using HMD-compatible 3D glasses and developed an abnormality determination process using optical flow techniques for detecting eye movements. The Gunnar Farneback method in OpenCV was used for optical flow detection due to its robustness to brightness variations.
2：Sample Selection and Data Sources:
22 participants were used in verification experiments, with results compared to conventional manual cover tests and Maddox rod tests as ground truth.
3：List of Experimental Equipment and Materials:
NVIDIA 3D Vision 2 glasses, web cameras, Arduino microcomputer, OpenCV software, illumination sources (663 lx and 896 lx tested).
4：Experimental Procedures and Operational Workflow:
The cover-uncover function was automated by controlling the liquid crystal shutters of 3D glasses via Arduino signals. Ocular images were captured continuously at 0.2 s intervals for 7-8 s after covering or uncovering an eye. Optical flow vectors were detected from image pairs, and a triangular region of interest was defined for analysis to detect abnormal eye movements.
5：2 s intervals for 7-8 s after covering or uncovering an eye. Optical flow vectors were detected from image pairs, and a triangular region of interest was defined for analysis to detect abnormal eye movements.
Data Analysis Methods:
5. Data Analysis Methods: Performance was evaluated by comparing system results with manual cover test and Maddox rod results, calculating inconsistency rates. Optical flow vectors were analyzed to determine abnormalities based on horizontal movement exceeding average flow magnitudes.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
