Development of a computerized system for analyzing facial nerve paralysis -A method using the optical flow-

我们开发了一个计算机系统来评估面神经麻痹的程度。我们的系统使用梯度法计算整个面部区域的光流。该系统测定结果精确，重现性好。它们几乎与使用匹配方法获得的结果相同，并且审查员之间的可变性最小。此外，当使用光流的疾病与健康侧比表示时，结果与使用Yanagihara分级量表分配给每个面部运动的分数具有良好的相关性。我们使用结果定义的光流指数也与柳原分级量表的总分高度相关。该系统的另一个优点是允许分析结果的灵活表达。例如，面部区域内每个小区域的光流可以叠加在患者的原始图像上。这使人们能够直观地理解面部运动。此外，面部区域中每个像素的运动可以在逐帧的基础上进行跟踪。这两个特征对于评估相关动作非常有用。我们的系统最初只在Linux上运行，但现在也可以在Windows上运行。期望该系统能在临床实践中得到广泛应用。

We developed a computerized system to assess the degree of facial nerve paralysis. Our system calculates the optical flow throughout the facial area, using the gradient method.The results obtained with our system were both precise and reproducible. They were almost identical to those obtained using the matching method, and the inter-examiner variability was minimal. In addition, when expressed using the diseased-to-healthy side ratio of the optical flow, the results correlated well with the scores assigned to each facial movement using Yanagihara's grading scale. The optical flow index, which we defined using the results, is also highly correlated with the total scores on Yanagihara's grading scale.Another advantage of our system is that it permits flexible expression of the analysis results. For example, the optical flow for each small area within the facial region can be superimposed on the original picture of the patient. This enables intuitive understanding of the facial movements. Furthermore, the movement of each pixel in the facial region can be traced on a frame-to-frame basis. These two features are very useful for assessing associated movements.Our system initially ran only on Linux, but it now also works on Windows. We expect that our system will be used widely in daily clinical practice.