三支决策扩展模型及其在荧光分子断层成像中的应用研究

Fluorescence molecular tomography (FMT) is a hot topic in molecular imaging that offers the possibilities to non-invasively monitor physiological and pathological processes at cellular and molecular level in vivo. But how to get the boundary of fluorescence target has always been a challenge in FMT. As an induced theory of rough sets, three-way decisions divide a universe into three pair-wise disjoint regions: positive (fluorescent target), neutral (three-dimensional boundary of target) and negative (background), which provides a possible way for this difficult problem. This project intends to construct generalized three-way decisions models and study their applications in fluorescence molecular tomography. The main contents are as follows. 1. This proposal is going to establish the theory of generalized three-way decisions model over two universes and three-way decisions model based on subset evaluation. We will focus on studying the mixed three-way decisions with the above two generalized models. 2. The project intends to develop a reconstruction frame for FMT based on the mixed three-way decisions model. First, we have to establish the two universes of FMT problem and build subsets of two universes. Then, we are going to design the evaluation functions based on the subsets and determine the thresholds. Combined with the mesh subdivision strategy, a high-resolution of target boundary will be delineated. This project may help promote the development of three-way decisions and expand its application. It will also promote the practicability of fluorescence molecular tomography. Consequently, this project will have important significance and application prospect.

荧光分子断层成像可在细胞和分子水平上对生物体生理病理变化进行在体监测，是分子影像研究的热点。但如何勾画荧光目标的三维边界一直是荧光分子断层成像的难题。三支决策作为粗糙集理论的延伸，由概率方法导出决策正域(荧光目标)、决策边界域(三维边界)和决策负域(背景区域)的思想，为解决这个难题提供了思路。本项目拟研究三支决策扩展模型理论及其在荧光分子断层成像中的应用，主要内容有：1.研究三支决策扩展模型，建立基于双论域的三支决策模型，构建基于子集评价的三支决策模型，重点研究基于双论域和子集评价的混合三支决策理论框架；2.研究基于混合三支决策的荧光目标重建框架，具体为建立重建问题的双论域，设计基于子集的评价函数和确定阈值，结合网格细分策略，勾画出高分辨率的目标三维边界。本项目将推动三支决策理论的发展，拓展其应用空间，推进荧光分子断层成像的实用化进程，具有重要的理论意义和应用前景。

本项目主要研究了三支决策扩展模型及其性质，并将其应用到荧光分子断层成像上，主要的研究成果及进展如下：.（1）研究了多种三支决策扩展模型及其性质，分别提出了基于子集评价的三支决策模型、基于不完备信息系统的三支模型、基于加权信息熵的直觉模糊信息系统的三支决策等，并研究其各种双论域模型；（2）将三支决策应用到荧光分子断层成像后处理，去除了伪影，仿真实验和真实实验都证明了基于三支决策的后处理策略可以提高荧光分子断层成像的重建结果质量；（3）开展了光学分子断层成像的快速重建算法研究，提出了基于传统模式识别方法的重建算法，同时也尝试开展基于深度学习的神经网络方法应用于光学分子断层成像重建。上述研究成果已在本领域国内外期刊上发表论文十几篇，申请专利3项。积极参与国内外同行举办学术会议5次。这些工作不仅对三支决策理论研究起到了积极的推动作用，同时也拓展了它在荧光分子断层成像中的应用，推动了光学分子成像的发展。