基于分数阶变分光流模型的左心室运动分析方法研究

Motion analysis of left ventricle (LV) is very important in clinical diagnosis and precaution of heart diseases. Using the theories of the digital image processing, fractional calculus, variational optical flow estimation and pattern recognition, a new LV motion analysis method based on fractional order variational optical flow (FOVOF) model is proposed to deal with the problems of the existing methods, and to improve their performances. The proposed method can be used to process real-time three-dimensional echocardiographs (RT3DE). The main research project is consist of the following three parts: .(1) The new methods for modeling different components of RT3DE are proposed by using the function space image modeling methods and fractional calculus. Based on this, the spots noise removal algorithms are proposed to improve the RT3DE quality. .(2) Combining the variational optical flow estimation and fractional calculus, a new FOVOF model is proposed to resolve the problems of the occlusion, large displacement and non-rigid motion, and get the LV's three-dimensional optical flow. .(3) Combining the feature selection and ensemble classifiers, a new disease assessment method based on the three-dimensional optical flow, which is helpful for the clinical doctors to analyze the motion of LV accurately, is proposed. .In theoretical research, the proposed motion analysis method can be used to analyze the motion of some non-rigid objects (right ventricle, lung, blood vessel and so on) and provides a new idea for the motion analysis of the non-rigid objects, therefore, it has important theoretical value and good prospect. In the application part, a diagnostic system of myocardial infarction (MI) will be developed to assess the performance of the proposed motion analysis method.

左心室运动分析在心脏疾病的诊断及预防中具有重要意义。本课题结合数字图像处理、分数阶微积分、函数空间图像建模、变分光流、模式识别等理论，提出一种基于分数阶变分光流模型的左心室运动分析方法，解决现有分析方法的不足，改善分析方法的性能。该方法适用于处理实时三维超声心动图（RT3DE），主要内容包括：① 探索对RT3DE中不同成分的新型建模方法，并据此提出斑点噪声去除算法改善RT3DE的质量；② 提出分数阶变分光流模型解决RT3DE中的遮挡、大位移和非刚性运动问题，得到左心室的光流场；③ 提出基于光流场的疾病评估方法，为医生正确分析左心室的运动情况提供辅助手段。在理论方面，该方法能用于解决一些非刚性物体的运动分析问题（如右心室、肺、血管等），为非刚性物体运动分析提供了新思路，具有重要的理论价值和发展前景；在实际应用方面，本课题将建立心肌梗死病情诊断系统，对方法的应用效果进行验证。

本课题结合数字图像处理、分数阶微积分、函数空间图像建模、变分光流等理论，针对基于分数阶微分理论的图像处理技术展开了深入研究。主要内容包括：（1）提出了鲁棒分数阶微分器的设计与实现方法，为研究基于分数阶微分理论的图像处理技术提供了基础保障；（2）从理论上阐述了分数阶微分算子的频域特性，据此提出了自适应分数阶图像增强算法，系统地论述了分数阶微分掩模算子对图像细节信息的增强作用；（3）提出了分数阶图像边缘检测和分割算法，有效地解决了整数阶算法的局限性，为图像边缘检测与分割提供了新工具；（4）构建了分数阶全变分去噪模型，并提出了多种求解算法，有效地解决了整数阶模型产生的阶梯效应问题，得到了更好的去噪效果；（5）提出了TV-L2盲去模糊算法，能够有效地消除由于运动而产生的模糊效果；（6）提出了基于分数阶变分光流模型的运动估计算法，为解决非刚性物体的运动估计问题提供了新思路。项目所产生的研究成果基本上形成了基于分数阶微分理论的图像处理技术的理论框架，具有较好的理论指导作用和应用推广价值。