OCT and OCTA image processing for retinal assessment of people with MS

用于多发性硬化症患者视网膜评估的 OCT 和 OCTA 图像处理

• 批准号: 10357873
• 负责人: Jerry L Prince
• 金额: $ 44.1万
• 依托单位: JOHNS HOPKINS UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-03-01 至 2025-02-28
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10357873
• 关键词: Address; Affect; Algorithmic Analysis; Algorithms; Anatomy; Angiography; Bayesian Modeling; Behavior; Biological; Biological Markers; Central Nervous System Diseases; Clinic; Clinical; Communities; Computational algorithm; Computer software; Computers; Data; Development; Diagnosis; Disease; Disease Management; Disease Progression; Endothelium; Equilibrium; Esthesia; Evaluation; Event; Eye; Eye diseases; Failure; Foundations; Four-dimensional; Future; Grant; Graph; Image; Imaging Device; Individual; Language; Learning; Literature; Longitudinal Studies; Machine Learning; Magnetic Resonance Imaging; Measurement; Measures; Methodology; Methods; Modeling; Monitor; Morphologic artifacts; Multiple Sclerosis; Muscular Atrophy; Neurodegenerative Disorders; Optical Coherence Tomography; Persons; Play; Process; Relapse; Research; Retina; Role; Sampling; Scanning; Severities; Source; Spinal Cord Lesions; Technology; Testing; Text; Thick; Thinness; Time; Training; Variant; Vision; Visit; advanced disease; automated algorithm; base; burden of illness; cerebral atrophy; cohort; convolutional neural network; cost; deep learning; density; diagnostic tool; disability; efficacy evaluation; experience; ganglion cell; gray matter; image processing; imaging Segmentation; imaging biomarker; imaging modality; improved; insight; learning progression; macula; macular edema; magnetic resonance imaging biomarker; nervous system disorder; novel; novel diagnostics; open source; portability; random forest; recursive neural network; research clinical testing; retina blood vessel structure; retinal imaging; segmentation algorithm; three dimensional structure; tool; treatment choice; white matter

Project Summary/Abstract (Max 30 lines of text) Although magnetic resonance imaging (MRI) is necessary for diagnosing multiple sclerosis (MS), it has been challenging to acquire consistent MRI measurements of MS disease burden. Spectral domain optical coherence tomography (OCT) of the retina has emerged as a complementary source of imaging biomarkers, wherein retinal thickness measurements have been shown to correlate well with MS disease burden. As well, OCT angiography (OCTA)—a new imaging modality acquired with the same OCT scanner—yields multiple new biomarkers among which macular vessel density has been shown to correlate with MS disability. There is strong evidence that OCT and OCTA may provide much needed imaging biomarkers for MS, but there are remaining technical challenges to overcome. Many algorithms for computation of retinal layer thicknesses from OCT images have been developed, but measurement of longitudinal changes in individual MS subjects remains highly challenging, especially in MS where yearly changes are small relative to intrinsic measurement variations. We propose a novel iterative registration and deep learning segmentation algorithm for longitudinal OCT retinal image segmentation. Development of automatic algorithms for analysis of OCTA images is in an early stage and there are opportunities for significant improvements. We will develop a deep network for OCTA vessel segmentation and biomarker computation that both suppresses artifacts that are common in OCTA and provides consistent results across different scanners. As both OCT and OCTA become more widely used in the characterization and management of MS, it is becoming increasingly important to jointly characterize these biomarkers and relate them to disease status, which is currently characterized largely by clinical evaluations. We will address the central question of whether OCT and OCTA can be used to predict disease progression by developing a new disease progression score for MS based on multiple OCT and OCTA measurements as well as clinical and MRI biomarkers, acquired in both single and multiple imaging visits. The proposed research will: 1) Develop a fast, topologically-correct longitudinal segmentation method for the macula; 2) Develop a method for artifact-suppressed and consistent computation of OCTA features in the macula; 3) Develop a disease progression score to jointly characterize longitudinal retinal OCT and OCTA measurements in MS; and 4) Carry out longitudinal studies of healthy controls and people with MS using OCT and OCTA measurements. We will assess whether average features within the macula or features averaged over smaller segments yield better estimates of progression. We will also assess whether OCT alone or OCT together with OCTA provide better estimates of progression. Image processing and disease progression algorithms will be made freely available to the research community. The proposed research will greatly advance the use of OCT and OCTA in characterizing longitudinal changes in the retina, potentially leading to standard eye measurements for monitoring and managing MS and other neurological and eye diseases.

项目摘要/摘要（最大30行文本） 尽管磁共振成像（MRI）对于诊断性多发性硬化症（MS）是必需的 挑战以获取MS疾病伯嫩MS疾病的一致测量。光谱域光学 视网膜的相干断层扫描（OCT）已成为成像生物标志物的完整来源， 其中已显示残留厚度测量与MS疾病伯嫩息息相关。也 OCT血管造影（OCTA） - 一种使用同一OCT扫描仪获得的新成像方式 - 黄斑血管密度中的新生物标志物已显示与MS残疾相关。有 有力的证据表明，Oct和Octa可能为MS提供急需的成像生物标志物，但是有 剩下的技术挑战要克服。许多用于计算残留层厚度的算法 已经开发了从OCT图像，但是测量了单个MS受试者的纵向变化 仍然受到高度挑战，尤其是在MS中，相对于固有测量，每年变化很小 变化。我们提出了一种新颖的迭代注册和纵向的深度学习分割算法 OCT视网膜图像分割。开发用于分析八章图像的自动算法在 早期，有重大改进的机会。我们将为八八 血管分割和生物标志物计算都抑制了在八八章中常见的伪影 在不同扫描仪之间提供一致的结果。随着OCT和OCTA都变得更广泛地使用 MS的表征和管理，共同表征这些越来越重要 生物标志物及其与疾病状况有关，目前主要由临床评估来表征。 我们将解决是否可以使用OCT和OCTA来预测疾病进展的核心问题 还基于多个OCT和八八测量，为MS开发新的疾病进展评分 作为临床和MRI生物标志物，在一次和多个成像访问中获得。拟议的研究将： 1）为黄斑开发快速的，拓扑校正的纵向分割方法； 2）开发一种方法 对于伪影抑制了黄斑中八八特征的一致计算； 3）发展疾病 在MS中的纵向残留OCT和八八体测量的共同表征的进展评分；和 4）使用OCT和OCTA测量进行对健康对照和MS的人进行纵向研究。 我们将评估黄斑内的平均特征还是在较小的细分市场中平均的特征 更好地估计进展。我们还将评估单独的十月还是与八八月一起评估 更好地估计进展。图像处理和疾病进展算法将自由 可用于研究社区。拟议的研究将大大提高OCT和OCTA的使用 在表征视网膜的纵向变化时，有可能导致标准的眼睛测量 监测和管理MS以及其他神经系统疾病。