基于深度学习的智能视力测量系统的应用实践  

Practical Application of Intelligent Vision MeasurementSystem Based on Deep Learning

作　　者：胡睿麟 孙丹 石桂莲 潘安鹏 HU Ruilin;SUN Dan;SHI Guilian;PAN Anpeng(National Clinical Research Center for Ocular Diseases,Wenzhou,325027;Eye Hospital,Wenzhou Medical University,Wenzhou,325027)

机构地区：[1]温州医科大学附属眼视光医院,温州市325027 [2]国家眼耳鼻喉疾病(眼部疾病)临床医学研究中心,温州市325027

出　　处：《中国医疗器械杂志》2024年第4期380-384,391,共6页Chinese Journal of Medical Instrumentation

基　　金：温州市基础性医疗卫生科技项目(Y2020919);浙江省自然科学基金探索公益项目(LTGY23H120001)。

摘　　要：为了全面评估临床干眼患者的真实视觉功能以及瞬目特征对人眼功能性视力的综合影响,设计开发了一种从侧面检测与分析受检者瞬目的智能视力测量系统。该系统基于深度学习关键点识别技术从侧面分析眼睑特征,对识别的上下眼睑关键点数据以折线图形式展示,并标记每次瞬目的波谷,通过基准值的设定,自动统计受检者完全瞬目与不完全瞬目的比例。结果表明,该系统性能稳定,测量准确,成功达到了预期的设计目标,可为未来的临床应用提供可靠的技术支持。To comprehensively assess the true visual function of clinical dry eye patients and the comprehensive impact of blinking characteristics on functional vision of the human eye,an intelligent vision measurement system has been designed and developed to detect and analyze blinks from the side.The system employs deep learning keypoint recognition technology to analyze eyelid features from a lateral perspective.It presents the data of identified key points for the upper and lower eyelids in a line chart format and annotates the trough of each blink.By setting benchmark values,the system automatically calculates the proportion of complete and incomplete blinks in the tested individuals.The results indicate that the system is stable in performance and accurate in measurement,successfully achieving the anticipated design objectives.It thereby provides reliable technical support for future clinical applications.

关 键 词：深度学习 侧面瞬目检测 视力波动 视觉功能 瞬目特征 

分 类 号：R197.39[医药卫生—卫生事业管理]