Web Camera Based Eye Tracking to Assess Visual Memory on a Visual Paired Comparison Task.
Web Camera Based Eye Tracking to Assess Visual Memory on a Visual Paired Comparison Task.
复制标题
DOI：
10.3389/fnins.2017.00370
复制
发表时间：
2017
影响因子：
4.3
通讯作者：
Zola S
中科院分区：
文献类型：
--
作者：
Bott NT;Lange A;Rentz D;Buffalo E;Clopton P;Zola S
背景：网络摄像头越来越成为大多数智能设备的标准硬件的一部分。眼球运动通常可以提供一个非侵入性的“大脑窗口”，使用网络摄像头记录眼球运动是一个新兴的研究领域。目的：本研究探讨了一种使用网络摄像头进行视觉配对比较（VPC）决策任务的新方法，为了进一步评估这种方法，我们检查了标准眼动跟踪摄像头自动评分程序[以每秒60帧（FPS）的速度获得图像]和使用内置笔记本电脑网络摄像头手动评分程序（以每秒3帧的速度获得图像）之间的相关性。研究方法：这是一项对54名临床正常老年人的观察性研究，受试者完成了三次门诊访视，同时通过标准眼动仪摄像头和内置笔记本电脑网络摄像头记录VPC决策任务的眼动。采用Siegel和Castellan的kappa公式分析评定者间信度。Pearson相关性用于研究使用标准眼动仪相机和内置网络相机的VPC性能之间的相关性。结果如下：在三次访视中的每次访视中，在60 FPS眼动仪和3 FPS内置网络摄像头之间的VPC平均新奇偏好评分上观察到强关联（r = 0.88-0.92）。各时间点网络摄像头评分的评分者间一致性较高（κ = 0.81-0.88）。VPC平均新奇偏好得分在10、5和3 FPS训练集之间有很强的相关性（r = 0.88-0.94）。使用内置的网络摄像头，数据质量问题明显减少。结论：使用内置笔记本电脑网络摄像头对VPC决策任务的人工评分与使用标准高帧率眼动仪摄像头对同一任务的自动评分密切相关。虽然这种方法不适用于需要收集和分析细粒度指标（如注视点）的眼动跟踪范例，但内置网络摄像头是大多数智能设备的标准功能（例如，膝上型计算机、平板计算机、智能电话），并且可以有效地用于以高精度和最小成本跟踪决策任务上的眼睛运动。
Background: Web cameras are increasingly part of the standard hardware of most smart devices. Eye movements can often provide a noninvasive “window on the brain,” and the recording of eye movements using web cameras is a burgeoning area of research. Objective: This study investigated a novel methodology for administering a visual paired comparison (VPC) decisional task using a web camera.To further assess this method, we examined the correlation between a standard eye-tracking camera automated scoring procedure [obtaining images at 60 frames per second (FPS)] and a manually scored procedure using a built-in laptop web camera (obtaining images at 3 FPS). Methods: This was an observational study of 54 clinically normal older adults.Subjects completed three in-clinic visits with simultaneous recording of eye movements on a VPC decision task by a standard eye tracker camera and a built-in laptop-based web camera. Inter-rater reliability was analyzed using Siegel and Castellan's kappa formula. Pearson correlations were used to investigate the correlation between VPC performance using a standard eye tracker camera and a built-in web camera. Results: Strong associations were observed on VPC mean novelty preference score between the 60 FPS eye tracker and 3 FPS built-in web camera at each of the three visits (r = 0.88–0.92). Inter-rater agreement of web camera scoring at each time point was high (κ = 0.81–0.88). There were strong relationships on VPC mean novelty preference score between 10, 5, and 3 FPS training sets (r = 0.88–0.94). Significantly fewer data quality issues were encountered using the built-in web camera. Conclusions: Human scoring of a VPC decisional task using a built-in laptop web camera correlated strongly with automated scoring of the same task using a standard high frame rate eye tracker camera.While this method is not suitable for eye tracking paradigms requiring the collection and analysis of fine-grained metrics, such as fixation points, built-in web cameras are a standard feature of most smart devices (e.g., laptops, tablets, smart phones) and can be effectively employed to track eye movements on decisional tasks with high accuracy and minimal cost.