IVOCT中冠状动脉分叉病变斑块分析及术后支架量化评估的自动算法研究

Bifurcation lesions, accounting for about 20% of the PCI cases and expecting to further increase, are continuous one of the most complex anatomic structures and remain a challenge for interventionalists. Without a single preferred treatment approach, a bifurcation lesion should be analyzed prior to the PCI to plan the interventional strategy, because bifurcation stenting often suffers from higher risks of acute and chronic complications such as acute thrombosis or late restenosis. As a novel image modality, intravascular optical coherence tomography (IVOCT) has played a particular role in the setting of contemporary stenting by providing very high resolution (10 μm) intracoronary images. Currently, most of the analysis in IVOCT images depends on the manual tracing and assessment, which is very time-consuming. .Due to the large amount of dataset that IVOCT can generate, scientific research based on IVOCT images is only feasible when the automated method is available. This study is intend to develop automated algorithms for the quantification analysis of bifurcation lesions and the implanted stents in IVOCT images. First, the side branches are automated located and the plaques in the bifurcation regions are detected and classified, using a machine learning based image pixel classifier. Next, the metallic stent struts and bioresorbable vascular scaffold struts are detected automatically for further processing. In the end, the skeleton of the implanted stent is automated reconstructed from the detected struts. This study will make it feasible to analysis a large dataset of IVOCT images efficiently. Besides, the automated methods could avoid the inter/intra-observer differences. The quantification analysis can provide the information like the types and distribution of plaques, the location of stents and even the degree of support to the vessel and blockage to side branches. Therefore, this study may contribute to a better understanding of bifurcation lesions, including the lesion classification, treatment optimization and PCI validation.

冠状动脉分叉病变是当前冠心病治疗中最为复杂，失败率最高的情况之一。研究证实，对病变斑块和支架的精细量化分析有利于此类病变的诊断、治疗和术后评估。血管内光学断层相干扫描（IVOCT）因其高分辨率的成像能力，愈来愈多的被用于术前冠脉分叉病变分析和术后支架评估。然而当前IVOCT图像处理工作还依赖于手工分析，严重影响了医生的工作效率及分析准确率，也使的基于IVOCT图像的大数据科学研究难以开展。在本课题中，我们将研究（1）IVOCT中冠脉分叉病变斑块的自动分割和分类算法；（2）术后金属支架和生物可降解支架的自动检测算法；（3）支架三维结构的自动重建算法，从而实现冠脉分叉病变的自动化分析，以提高工作效率和准确性，减少手工分析所带来的观测者间差异和观测者自身差异。分析和评估结果可以进一步用于指导临床诊断、手术规划，分析支架的放置精度、支撑力和对分支的血流影响。因此本研究有着较为重要的科学价值。

由于其内在复杂的病变情况，冠状动脉分叉病变是最富有挑战，术后不良事件发生率最高的病症之一。为制定合理的治疗方案，医生需要关于冠状动脉分叉病变的位置、区域内斑块病变的分布和类型等精准信息。为了降低术后不良事件，还需要及时量化评估植入后的支架贴壁情况。血管内光学相干断层扫描（Intravascular Optical Coherence, IVOCT）是目前唯一可以提供以上全部信息的技术，分辨率高达10微米。如何在成百上千张图像中，快速、自动实现血管分支检测、斑块的检测和分类、支架贴壁情况的量化分析等，不但对手术方案选择、病人预后评估都有着至关重要的影响，也有助于我们更好的理解和研究冠脉分叉病变。.针对这些问题，我们首先制备了一系列均质仿体，探索了扫描光源位置区别对IVOCT图像的影响，提出了一组光衰减模型用于校正因为光源距离和入射角度不同造成图像灰度值差异，并基于回声公式增强了深层组织的信号强度，可以更好的展示出斑块病变。然后我们构建了一系列基于BP神经网络、支持向量机和随机森林的机器学习模型，实现了对富脂斑块、纤维斑块、钙化斑块和炎症富集等病变的自动检测和分类，测试显示其正确率分别为87.6%、89.6%、76.1%和81.2%，达到了国际领先水平。为了自动定位血管分支位置，我们完成了一种基于支持向量机和三维先验信息的自动算法，测试显示，血管分支的检出正确率达到94.3%，血管分支管腔截面检测结果的Dice指数为0.898，血管分支的误检率约为2.1%。在校正后的图像基础上，我们利用扫描线上灰度值的变化趋势规律，实现了对金属基支架的自动检测，并结合血管腔的位置，给出支架贴壁情况的评估值。.研究较好的达到了项目预期的目标，实现了冠脉分叉病变区域的斑块分析和术后支架的量化评估，可以为临床研究和治疗冠状动脉分叉病变提供更精准的量化信息。.