Validating a machine learning model of eye tracking in children with cortical visual impairment (CVI)

验证皮质视觉障碍 (CVI) 儿童眼球追踪的机器学习模型

• 批准号: 10595052
• 负责人: Melinda Chang
• 金额: $ 23.46万
• 依托单位: CHILDREN'S HOSPITAL OF LOS ANGELES
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-04-01 至 2027-03-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10595052
• 关键词: Ablation; Active Learning; Adopted; Adverse effects; Age; Behavior; Biomedical Engineering; Brain; California; Categories; Characteristics; Child; Childhood; Classification; Clinic; Clinical; Clinical Trials; Clinical Trials Design; Clinical assessments; Collaborations; Contrast Sensitivity; Cornea; Crowding; Data; Data Analyses; Data Science; Data Set; Developed Countries; Disease; Electrophysiology (science); Enrollment; Evaluation; Evidence based treatment; Exhibits; Eye; Foundations; Frequencies; Future; Goals; International; Interview; Joints; K-Series Research Career Programs; Knowledge; Laboratories; Learning; Light; Longitudinal Studies; Los Angeles; Machine Learning; Measures; Mentored Patient-Oriented Research Career Development Award; Mentors; Methods; Neural Network Simulation; Neurodevelopmental Disorder; Neurologic; Ophthalmology; Outcome Measure; Parents; Patients; Pediatric Hospitals; Protocols documentation; Quality of Life Assessment; Quality of life; Questionnaires; Research; Research Personnel; Resolution; Saccades; Sampling; Severities; Techniques; Technology; Time; Training; Translating; Universities; Vision; Vision Tests; Visual; Visual Acuity; Visual Pathways; Visual System; Visual evoked cortical potential; Visual impairment; cohort; comorbidity; computer monitor; cortical visual impairment; deep learning; deep neural network; experience; gaze; large datasets; machine learning model; neurodevelopment; next generation; novel; oculomotor; programs; prospective; sex; signal processing; success; targeted treatment; translational potential; visual control; visual dysfunction; visual stimulus; visual tracking

Project summary Cortical visual impairment (CVI) is the leading cause of pediatric visual impairment in developed countries. There is no evidence-based treatment, and design of clinical trials is hampered by the absence of a validated method of visual assessment that captures the numerous aspects of visual function that are compromised in pediatric CVI. Our laboratory is investigating the use of eye tracking in children with CVI. During eye tracking, an infrared camera tracks the pupillary and corneal light reflections while a child watches visual stimuli on a computer monitor. The eye tracker calculates the direction of eye gaze with high spatial and temporal frequency. Our eye tracking protocol assesses multiple afferent, efferent, and higher-order visual parameters during a 12-minute recording session. Our initial data show that eye tracking is reliable and quantifies multiple visual and oculomotor parameters in children with CVI. Given the large amount of data generated by eye tracking (2,000 data points per second), higher-level analytics are required. We will validate a machine-learning model of eye tracking in children with CVI via three Specific Aims. In Aim 1, we will quantify deficits of visual function in pediatric CVI using eye tracking, strengthening the findings in our preliminary data by inclusion of a well-powered sample. In Aim 2, we will use machine learning to develop a CVI eye tracking severity score. In Aim 3, we will validate eye tracking by comparing and contrasting with two other methods of visual assessment in children with CVI, sweep visual evoked potentials and the CVI Range. Together, these studies will establish eye tracking as a quantitative, objective, and comprehensive measure of visual function in pediatric CVI. In the R01 application planned at the end of the K23 award period, we will incorporate the CVI eye tracking severity score as an outcome measure in a longitudinal study of standard and targeted therapies for CVI. In pursuit of these aims, I will be mentored by a highly experienced, interdisciplinary, internationally recognized team at Children’s Hospital Los Angeles and University of Southern California. Under their guidance, I will also pursue a Masters degree in Applied Data Science and gain experiential learning in electrophysiology. The training acquired during my Career Development Award will enable me to transition to an independent investigator leading a research program focused on developing next-generation technologies to interrogate the visual system in children with a variety of neurodevelopmental disorders.

项目摘要 在发达国家，皮质性视力损害（CVI）是儿童视力损害的主要原因。那里 缺乏循证治疗，临床试验的设计因缺乏有效的方法而受到阻碍 视觉评估，捕捉儿童视觉功能受损的许多方面， CVI。我们的实验室正在研究眼动追踪在CVI儿童中的应用。在眼动追踪过程中， 当孩子在电脑上观看视觉刺激时，摄像机跟踪瞳孔和角膜的光反射 监视器.眼动仪以高的空间和时间频率计算眼睛注视的方向。我们的眼睛 跟踪协议评估多个传入，传出，和高阶视觉参数在12分钟内， 录音会议。我们的初步数据表明，眼动追踪是可靠的，并量化了多个视觉和眼动 儿童CVI的参数。考虑到眼动追踪产生的大量数据（2，000个数据点 需要更高级别的分析。我们将验证眼动跟踪的机器学习模型， 通过三个具体目标，儿童CVI。在目标1中，我们将量化儿童CVI的视功能缺陷 使用眼动追踪，通过纳入一个有效样本来加强我们初步数据中的发现。在 目标2，我们将使用机器学习来开发CVI眼动追踪严重程度评分。在目标3中，我们将验证眼睛 通过与其他两种视觉评估方法的比较和对比，跟踪儿童CVI，扫描 视觉诱发电位和CVI范围。总之，这些研究将建立眼动跟踪作为一个定量的， 客观、全面地评价儿童CVI的视功能。在R01应用计划中， 在K23奖励期结束时，我们将把CVI眼动追踪严重程度评分作为一项结果指标， CVI标准和靶向治疗的纵向研究。为了实现这些目标，我将接受一位 洛杉矶儿童医院和洛杉矶儿童医院的经验丰富、跨学科、国际公认的团队 南加州大学。在他们的指导下，我还将攻读应用数据硕士学位 科学和获得电生理学的经验学习。在我的职业生涯中获得的培训 发展奖将使我能够过渡到一个独立的调查员领导的研究计划 专注于开发下一代技术，以询问儿童的视觉系统， 神经发育障碍