The height of the lacrimal river can reflect the secretion of the lacrimal gland and is one of the important indicators for evaluating the condition of patients with dry eye. Common tear meniscus height detection methods include traditional methods based on digital image processing technology and manual measurement based on tear meniscus images. The former is easily affected by eye tissue interference, image shooting angle, blinking and other problems, while the latter will reduce work efficiency due to time and labor costs. In view of the above problems, this paper proposes a deep learning-based method for detecting the height of the tear meniscus. Specifically, for the acquired human eye images, the open and closed eyes classification network based on convolutional neural network, the coarse location network of the tear meniscus region, the tear meniscus fast segmentation network, the tear meniscus edge correction and the tear meniscus height calculation module are successively used to realize The data processing of human eye images automatically detects the height of the tear meniscus. The method in this paper obtains 99.5% eye opening recognition accuracy, 99.1% eye opening recall, and 91.6% tear meniscus area intersection ratio and other results in the experiment. The experimental results show that, compared with the traditional digital image processing method, the proposed method has stronger robustness and more accurate detection, and can also replace manual measurement and improve work efficiency.
Reference: Wang Chongyang, Chen Wenguang. Tear meniscus height detection method based on deep learning [ J ]. Computer Application Abstracts, 2022, 38 ( 20 ) : 5.