报告题目I：Non-parametric statistical methods to compare the areas under correlated LROC curves

报告时间：2015年11月18日（周三），10:00 - 11:30

报告题目II：Influence of the grayscale on early assessment of image quality in X-ray computed tomography   

报告时间：2015年11月18日（周三），14:00 - 15:30

报告摘要：

Nonparametric statistical methods to compare the areas under correlated LROC curves

Task-based image quality assessment is essential for the development, optimization and evaluation of new image formation processes in X-ray computed tomography, as well as in other imaging modalities. ROC analysis with signal localization, which is referred to as LROC analysis, is a rich approach to perform such an assessment. It provides a means to jointly assess the accuracy of visual search and detection in an observer study. Reader availability, reader fatigue, and limited computer resources usually limit the amount of images involved in an LROC study, so that the interpretation of results require a strong understanding of statistical variability. We have recently developed two useful methodologies to assess this variability in the context of early assessment of imaging systems with fixed reader effects. The first methodology applies to LROC studies that are conducted in such a way that the entire LROC curve is first obtained, with the area under it computed in a second step, by integration. The second methodology applies to LROC studies that are conducted using alternative forced-choice experiments. Both methodologies will be explained in detail, and their usefulness will be illustrated in the context ofa comparison between reconstruction algorithms in CT.

Influence of the grayscale on early assessment of image quality in X-ray computed tomography

The radiation dose associated with CT scans has become an important concern lately. To ensure that the risks associated with CT do not outweigh the benefits, it is essential to reduce the dose imparted to the patient by CT scans. A promising approach to reduce dose is the utilization of non-linear reconstruction algorithms.Early-stage assessment of image quality achieved using such algorithms is a challenging problem, and this problem is further complicated by the need to compare a large variety of options corresponding to different parametric settings. We will present here a study that aims at understanding better how the grayscale window affects human observer performance in this context. The results of this study imply that close attention should be given to the grayscale window when comparing non-linear iterative reconstruction algorithms. This implication has been verified with another study, the results of which will also be presented, along with a mathematical analysis of the grayscale window effect on the statistics of the image.

报告人简介：Frederic Noo 教授，1998年于比利时 University of Liege获得医学成像博士学位，自2002年加入University of Utah，2013年被评为Tenure教授。曾多次担任Medical Physics的客座编辑，图像重建领域内最重要的学术会议International Meeting on Fully Three-dimensional Image Reconstruction in Radiology and Nuclear Medicine和International Conference on Image formation in X-ray Computed Tomography的主席和会议论文主编。目前担任IEEE Tran Medical Imaging的副编辑，Medical Physics 的编委会委员。曾任犹他大学先进成像研究和设施基金委员会主席，终身教授评审委员会委员，放射系执行委员会成员等。他在CT成像领域的重建方法和性能评估研究方面有深厚的研究经验和突出的研究成果，尤其在解析重建方面发表了多项代表性工作。

第147期“论坛”主请人联系方式：
邢宇翔   62782510   xingyx@tsinghua.edu.cn