多视图学习及其在医学图像分析中的应用研究

本项目将研究多视图特征选择、多视图字典学习的若干模型和相关算法，同时将会研究多视图快速凸优化方法来解决大规模多视图学习问题。此外，我们将所提方法应用于三个具体临床辅助诊断问题（1）多视图MR前列腺图像分割，（2）近红外前哨淋巴结定位，（3）肺癌病理学细胞图像识别。 基于所研究成果，将在重要或者有影响的国际学术刊物以及国际会议上发表论文6-8篇，申请专利3项。通过本项目的研究，将进一步加强多视图学习框架的理论深度和应用广度。

This project is going to study the multi-view feature selection, multi-view dictionary learning,as well as the multi-view convex optimization for large-scale multi-view learning problem. Also, we will apply the proposed methods to three particular medical image analysis applications (i) multi-view MR prostate segmentation, (ii) NIR-based sentinel lymph node localization, and (iii) lung cancer image identification based on needle biopsy specimens. It is expected to publish 6-8 articles in top journals and conferences as well as 3 patents applications.The project aims to enhance the aspect of both the theory and the technologies for multi-view learning.