Retrospective Motion Compensation in Optical Coherence Tomography

光学相干层析成像(OCT)为非侵入性、活体、微米级的散射组织(如视网膜)进行2D和3D成像提供了可能。使用3D-OCT，可以在眼科进行定量测量，有助于及早发现和治疗青光眼和糖尿病视网膜病变等眼部疾病。3D-OCT数据集不是即时获取的，而是由几秒钟内记录的许多一维轴向扫描组成的。因此，数据集显示了扫描过程中眼睛运动导致的失真和伪影。OCT图像也包含斑点噪声。我们通过使用一个区域的多个3D扫描来解决这些问题，这些区域彼此配准以校正运动，并被组合以减少斑点噪声。我们利用OCT成像过程的时序结构，确定轴向扫描之间的预期运动量，从而确定运动造成的失真。使用不同扫描模式获取的OCT图像用于回溯估计和补偿对象运动。提出了一种新的纯软件配准方法，该方法基于全局的、特定于问题的目标函数的优化，该目标函数能够在所有三个维度上校正运动。结合用于去噪和图像合并的特殊后处理，该方法在获得更准确的数据和改善图像质量方面具有巨大的潜力，而不会增加基于硬件的运动校正方法的额外成本和复杂性。准确和重复地对视网膜进行成像的能力可以提高对疾病的早期、准确诊断和治疗监测。该项目将高级配准和图像增强方面的专业知识转移到OCT成像，旨在开发新的解决方案，以满足基本的临床需求。

Optical coherence tomography (OCT) opens up the possibility to perform non-invasive, in vivo, micron scale 2D and 3D imaging of scattering tissue, such as the retina. Using 3D-OCT, quantitative measurements can be performed in ophthalmology, facilitating early detection and treatment of eye diseases, such as glaucoma and diabetic retinopathy. 3D-OCT datasets are not acquired instantaneously but are composed of many 1D axial scans recorded within few seconds. Thus, datasets show distortions and artefacts that result from motion of the eye during the scan. OCT images also contain speckle noise. We approach these problems by using multiple 3D scans of an area which are registered with each other to correct for motion and are combined so that speckle noise is reduced. We utilize the timing structure of the OCT imaging process, which determines the expected amount of motion between axial scans and therefore the distortion caused by motion. OCT images that were acquired with different scanning patterns are used to retrospectively estimate and compensate object motion. A novel software only registration method based on the optimization of a global, problem-specific objective function that is able to correct motion in all three dimensions is proposed. Combined with special post-processing for de-noising and image combination this approach offers great potential for achieving both more accurate data and improved image quality without the additional cost and complexity of hardware based motion correction methods. The ability to accurately and repeatably image the retina could improve the early, accurate diagnosis of disease and monitoring of treatment. The project transfers expertise in advanced registration and image enhancement to OCT imaging and aims to develop novel solutions to meet basic clinical needs.