【作者】 陈远；

【导师】 赵志敏；

【作者基本信息】 南京航空航天大学 ， 精密仪器及机械， 2012， 博士

【摘要】 微循环是反映人体健康的重要指标，在许多生理和病理状态下有着重要的参考价值，对其进行可靠有效的自动检测，在预防疾病的发生和药物疗效检测方面有着极其重要的作用。在过去微循环的测量由人工完成，这不仅花费大量人力物力，而且无法保证测量精度，为了克服这些缺点，一些研究人员使用计算机技术开展了微循环的自动化测量研究。虽然基于计算机的微循环自动化测量在一些方面取得了较大的进展，但依然有许多问题有待进一步研究。为了促进微循环学科的发展，推进微循环的计算机自动测量与识别工作，本文进行了以下几个方面的研究：1、针对光学无损方式采集的微循环图像具有光照不均、噪声大、对比度低等特点，研究了基于多尺度分析的光照不均匀图像的校正方法，提出了基于Contourlet变换的微循环图像增强方法，实现了微循环图像的增强，改善了微循环图像的可视化效果。2、为了实现微循环血管形态参数的自动测量，提出了基于视频图像序列的微循环参数自动测量方法，该方法实现了微循环形态参数的自动测量。研究了有血管区域和无血管区域的判断准则；设计了血管中心线自动迭代提取算法，包括种子点的自动选取、中心线迭代算法的设计以及停止条件判断准则；研究了血管边界和直径以及血管曲率的自动测量方法。分析了测量参数对微循环病变识别的统计特性。3、针对不同类型微循环血管流态的复杂性，研究了基于ST图的流速测量方法。提出了基于正交滤波器组测量微循环流速的方法。比较了提出方法与现有方法的测量精度和测量鲁棒性，讨论了各种方法测量失败的原因，结果验证了提出方法的优越性。4、为了实现微循环的血液细胞跟踪与测量，提出了基于ST图的细胞跟踪与测量方法。设计了基于多尺度方向滤波器的ST图像增强方法，设计了噪声滤波函数和方向滤波函数，设计了轨迹细化和分岔处理算法，实现了细胞的跟踪和流速测量。分析了滤波器参数的取值方案，讨论了单条跟踪线对细胞跟踪的局限性。5、针对单条跟踪线无法跟踪微循环血液细胞复杂运动的情况，研究了基于三条跟踪线的细胞跟踪方法。提出了多候选轨迹自动分组准则和多候选轨迹的融合准则，实现了细胞的自动跟踪。为了验证细胞跟踪的有效性，提出了细胞跟踪有效性验证准则。针对三条跟踪线无法跟踪较大血管细胞运动的情况，研究了多跟踪线的细胞跟踪方法。提出了多跟踪线的轨迹融合准则以及最优跟踪线数量的选择准则。讨论了多跟踪线的方向选择问题。6、为了实现微循环的计算机自动辅助诊断，研究了微循环的自动识别方法。在微循环血管参数自动测量的基础上，以测量参数作为特征向量，利用模式识别方法对微循环血管进行分类与识别，结果显示，本文研究的微循环自动识别方法能够有效地应用在疾病的辅助诊断中。更多还原

【Abstract】 The parameters of the microcirculation are the indictors for the health of the human, which shown theimportant reference value in pathological and physiological aspects. Automatic detecting andrecognizing the patterns of microcirculation securely and efficiently have important effect in preventingthe disease occurrence and the therapeutic drug monitoring. However, the traditional way for theparameters of microcirculation detection was via manual measurement, which was not onlytime-consuming, but also lack of measurement accuracy. To handle this shortcoming, some researchersdeveloped the techniques for automatic detection of the parameters of microcirculation by usingcomputer analysis.Although great advantage progress obtained in the automatic detection of the parameters ofmicrocirculation based on computer analysis, there are still lots of problems should be furtherresearched. In order to improve the development of microcirculation, and promote the research ofautomatic detection and recognition of microcirculation based on computer analysis, we have doneseveral studies in the following aspects.First, according to the fact that the microcirculation images captured by non-destructive opticalimaging method were usually with uneven illumination background, strong noise and low contrast, theillumination correction method was studied, and then the microcirculation images enhancement methodbased on Contourlet transform was proposed. The enhancement of microcirculation images wererealized, which improved the images visualization effect.Second, in order to realize the parameters of microcirculation automatic detection, the parameters ofmicrocirculation automatic detection method based on video frame sequence was developed. By usingthe developed method, the automatic measurements of geometric parameters were realized. Thedecision criteria for the area that with a vascular or not was proposed; the algorithm for automaticiterative extraction of blood vessels centerline was proposed, which including automatic selection ofthe seed points, the iterative algorithm design and stopping criteria establishment; the fast calculationmethod for the boundary, diameter and curvature of blood vessel were researched. The statisticalcharacters of the measured parameters for abnormal microcirculation recognition were analyzed.Third, aim at the complication of the measurement of blood flow velocity in different types ofmicrocirculation blood vessels, the method for velocity measurement based on ST image wasresearched. A new method for measurement of the flow velocity of microcirculation based on a quadrature filter banks was proposed. The precision and robustness of the proposed method were alsocompared with the existed methods. The reasons for the failure measurement were also discussed. Theresults showed the superiority of the proposed method.Fourth, in order to realized the automatic tracking and measurement of the motion of blood cells ofmicrocirculation, the method for blood cells tracking and measurement based on ST image wasproposed. The ST image enhancement method based on multiscale directional filters was designed, thenoise suppression function and orientation filtering function were designed, the traces thinning andbifurcation traces cutting algorithm were designed, and then the blood cells tracking and velocitymeasurement were realized. The way of the parameter selection of the filter was analyzed. Thelimitations that a single trackpath for blood cells tracking were discussed.Fifth, to resolve the helpless tracking of the complicated motion of blood cells in microcirculation byusing a single trackpath, a method for blood cells tracking based on three trackpaths was researched.The automatic grouping criteria of traces candidates was proposed, the fusion criteria of tracescandidates was proposed, which realized the automatic blood cells tracking. In order to verify thevalidation of the tracking results of blood cells, the validation verification criteria of blood cellstracking was proposed. For the situation that three trackpaths were not enough for tracking the motionof blood cells in larger blood vessel in microcirculation, a method based on multiple trackpaths wasresearched. The traces fusion criteria of multiple trackpaths was proposed and the criteria of optimalnumber of trackpaths selection was proposed. Finally, the direction of the multiple trackpaths selectionwas discussed.Sixth, in order to realize the computer aided diagnosis for microcirculation, the automated recognitionmethod was researched. After the parameters of microcirculation were measured automatically, theparemeters were formed as the feature vectors and the pattern recognition method were employed toautomatically classify and recognize the microcirculation. The results show the proposed method canbe applied in disease assistant diagnosis efficiently.更多还原