3D segmentation and registration of macular SD-OCT for application in MS

描述（由申请人提供）：至少55%的多发性硬化症（MS）患者视力受损，这可能是疾病发病的第一个表现。光谱域光学相干断层扫描（SD-OCT）使视网膜的体内高分辨率研究成为可能，并且越来越多地被用作神经退行性疾病的生物标志物。SD-OCT提供了视网膜神经纤维层、神经节细胞层和整体视网膜变薄的表型测量，表明多发性硬化症中轴突和神经元视网膜病理。有令人兴奋的证据表明，多发性硬化症中更深的视网膜层也受到影响，这是特别有趣的，因为这些神经元从未有髓鞘。因此，SD-OCT测量已被证明在MS病理生理学新科学理论的发展中是有用的。这些测量在疾病的早期检测、疾病严重程度分期、评估疾病进展和确定单个MS患者的治疗效果方面也有潜在的作用。尽管先进的自动分割和测量视网膜关键特征的算法在研究实验室和商业仪器中都出现了，但要充分利用这种三维成像技术，仍然存在几个关键的技术限制。首先，视网膜分割方法目前不能提供亚体素精度，也不能保证黄斑水肿的存在。其次，虽然视网膜的三维比较在标准眼科检查中是常规的，但所使用的黄斑定位方法并没有将刚性和可变形成分分开，以便在标准化空间中进行详细的总体比较。第三，视网膜厚度一般沿直线外部计算，不补偿相对姿态。总之，这些缺陷阻碍了三维纵向和横断面科学研究，并限制了对特定受试者疾病进展的监测。本研究将：1)开发一种抗水肿的黄斑亚体素视网膜层分割方法；2)开发刚性和可变形的配准方法，允许在标准化空间中分析SD-OCT体积；3)发展视网膜固有厚度测量技术和平均黄斑寰椎空间；4)在标准化空间中对正常受试者和MS患者进行横断面和纵向研究，以检验i)更深的视网膜层参与的假设

DESCRIPTION (provided by applicant): Vision is compromised in at least 55% of multiple sclerosis (MS) patients and may represent the very first manifestation of disease onset. Spectral domain optical coherence tomography (SD-OCT) enables in-vivo, high-resolution studies of the retina, and is increasingly being used as a biomarker in neurodegenerative diseases. SD-OCT has provided phenotypical measurements of the retinal nerve fiber layer, ganglion cell layer, and overall retinal thinning, indicating both axonal and neuronal retinal pathology in MS. There is tantalizing evidence that deeper retinal layers are also affected in MS, which is of particular interest since these neurons are never myelinated. Thus, SD-OCT measurements have proven useful in the development of new scientific theories about the pathophysiology of MS. These measurements are also potentially useful in early detection of disease, staging disease severity, assessing disease progression, and determining therapeutic efficacy for individual MS patients. Although advanced automated algorithms for segmentation and measurement of key retinal features are emerging in both research labs and commercial instruments, there remain several key technical limitations to full exploitation of this three-dimensional imaging technique. First, retinal segmentation methods do not presently provide subvoxel precision nor are they robust to the presence of macular edema. Second, although three-dimensional comparisons of retinas are routine in standard ophthalmologic exams, the macular registration methods that are used do not separate the rigid and deformable components for detailed population comparisons in a normalized space. Third, retinal thicknesses are generally computed extrinsically along straight lines without compensation for relative pose. Together, these deficiencies hamper 3D longitudinal and cross-sectional scientific studies and limit monitoring disease progression in specific subjects. The proposed research will: 1) Develop a subvoxel retinal layer segmentation method for the macula that is robust to edema; 2) Develop both rigid and deformable registration methods that will permit analysis of SD-OCT volumes in a normalized space; 3) Develop intrinsic retinal thickness measurement techniques and an average macular atlas space; and 4) Carry out both cross-sectional and longitudinal studies of normal subjects and MS patients in a normalized space to test the hypotheses that i) deeper retinal layers are involved in MS and ii) longitudinal atrophy is observed in deeper retinal layers in MS. We will also explore whether regional macular edema precedes thinning in the inner nuclear layer in MS subjects. These studies will also include function/structure regression using the full macular volume, longitudinally assessed, against visual acuity functional scores. Image processing and regression algorithms will be developed within the open-source Java Image Science Toolkit (JIST) framework and released as open source software.