AN AUTOMATED IMAGE ANALYSIS AND MEASUREMENT SYSTEM FOR VIDEO-FLUOROSCOPIC EVALUATION OF SWALLOWING DYSFUNCTIONS.

虽然吞咽是健康的人每天毫不费力地执行数百次的功能，但这是一个极其复杂的过程，涉及人体内大量肌肉和组织的快速而精确的协调。在中风(在英国，每年有数千人受到影响，NHS每年花费23亿英镑)和许多其他神经系统疾病中，吞咽困难(异常吞咽)经常发生，可能导致致命的肺炎。视频透视(VF)是诊断吞咽困难的金标准。对于VF吞咽研究，患者舒适地坐着，并给予浓缩的钡(使其在X射线中变得不透明)液体。病人吞咽时，头部和颈部的X光视频图像正在被记录下来。这段视频必须由临床医生手动检查，但这项任务对视觉要求很高，非常耗时，而且容易出错，因为它需要以慢动作逐帧重放整个视频，并非常仔细地检查涉及的解剖区域。这对诊断的准确性有直接影响。我们建议开发一个系统，自动处理整个视频序列，并计算对临床医生稳健诊断至关重要的测量结果。具体地说，该系统将能够自动完成这项工作，并为临床医生提供所需的测量，而不是临床医生以慢动作逐帧检查视频数据。我们建议通过开发图像处理算法来实现这一点，该算法将图像分割为其组成区域。这些区域将是特定的解剖区域和吞咽时的液体。通过在一段时间内自动跟踪所有这些区域，系统将能够计算临床医生执行诊断所需的所有测量。在算法开发完成后，我们计划将该系统应用于正常和受影响的受试者，并将自动计算的测量结果与手动估计的测量结果进行比较。这项拟议的工作将对患者、NHS和科学界产生重大好处。这是新颖的，因为以前的工作都没有尝试所提出的自动化，并且由于当前需要提高基于VF的评估的有效性，所以是及时的。

Although swallowing is a function performed effortlessly hundreds of times daily by healthy humans, it is an extremely complex process that involves the rapid and precise coordination of numerous muscles and tissues in the human body. In stroke (that affects thousands of people every year in the UK with an annual NHS cost of 2.3 billion) and many other pathologies of the nervous system, dysphagia (abnormal swallowing) frequently occurs and can lead to fatal pneumonia. Video-Fluoroscopy (VF) is the gold-standard for diagnosing dysphagia. For a VF swallowing study, the patient is seated comfortably and given barium enriched (to become opaque in the x-ray) liquid. The patient swallows while x-ray video images of the head and neck are being recorded. This video has to be manually examined by clinicians, but this task is visually demanding, extremely time consuming and error prone, since it requires the replay of the entire video frame-by-frame in slow motion and the very careful examination of the involved anatomical areas. This has direct consequences to the diagnostic accuracy.We propose to develop a system that processes the entire video sequence automatically and calculates measurements that are critical for robust diagnosis by the clinician. Specifically, instead of the clinicians examining the video data frame-by-frame in slow motion, the system will be able to do the job automatically and provide the clinicians with the required measurements. We propose to do this through the development of image processing algorithms that segment an image to its constituent regions. These regions will be the specified anatomical areas and the liquid during swallowing. By automatically tracking all these regions during time, the system will be able to calculate all measurements needed by the clinician to perform a diagnosis. After the algorithms are developed, we plan to apply the system to normal and affected subjects, and compare the automatically calculated measurements with ones estimated manually. The proposed work will have significant benefits on patients, NHS and the scientific community. It is novel as none of the previous works attempted the proposed automation, and is timely due to the current needs for improving the effectiveness of VF-based evaluation.