冠心病可以通过评估左心室超声图像中心壁的区域运动来诊断。
Coronary Heart Disease can be diagnosed byassessing the regional motion of the heart walls in ultrasound images of theleft ventricle.
即使是专家也难以准确解释超声图像，从而导致不同观察者之间的判断发生变化。
Even for experts, ultrasound images aredifficult to interpret leading to high intra-observer variability.
以前的研究工作表明，为了解决这个问题，需要考虑不同心脏区域之间的相互作用及其对心脏临床状况的总体影响。
Previous work indicates that in order toapproach this problem, the interactions between the different heart regions andtheir overall influence on the clinical condition of the heart need to beconsidered.
为此，我们提出了一种联合学习条件随机场结构和参数的方法，将这些任务描述为一个凸优化问题。
To do this, we propose a method for jointlylearning the structure and parameters of conditional random fields, formulatingthese tasks as a convex optimization problem.
我们考虑每个与边缘相关的特征集的块-L1正则化，并构建一个有效的投影方法来找到全局最优的惩罚最大似然解。
We consider block-L1 regularization foreach set of features associated with an edge, and formalize an efficientprojection method to find the globally optimal penalized maximum likelihoodsolution.
与本文相关的网站供参考：
https://www.cs.ubc.ca/~murphyk/Software/L1CRF/index.html
下载英文原文地址：
http://page2.dfpan.com/fs/4lfcaj8212b17219166/
更多精彩文章请关注微信号：
