论文题名(中文): | 基于C扫描图的视盘OCT图像运动伪差校正算法的研究 |
姓名: | |
论文语种: | chi |
学位: | 硕士 |
学位类型: | 学术学位 |
学校: | 北京协和医学院 |
院系: | |
专业: | |
指导教师姓名: | |
校内导师组成员姓名(逗号分隔): | |
论文完成日期: | 2016-05-25 |
论文题名(外文): | Study of Motion Artifacts Correction Algorithm in Optical Coherence Tomography Images Based on C-Scan of Optic Disc |
关键词(中文): | |
关键词(外文): | OCT motion artifact X direction artifact correcting repetition and missing |
论文文摘(中文): |
光学相干层析成像(Optical Coherence Tomography, OCT)技术作为一种新型无损光学成像技术,可对组织的形态、内部结构进行二维成像和三维可视化。具有灵敏度高、分辨率高、实时活体成像、非接触性、横向分辨率和纵向分辨率相互独立等优点。OCT技术最重要的医学贡献在眼科,凭借其能突破传统眼科对视网膜断层结构和功能成像,成为诊断眼科疾病的标准工具。 OCT系统在进行人体活检的三维成像时,由于眼球的不自主运动可能会导致采集的图像体数据畸变和错位,使得到的体数据失去真实性,生理结构得不到真实反映,从而可能产生临床诊断的误诊、漏诊。为解决这一问题,相关实验室提出了校正X方向和Z方向的畸变的方法。这些研究方法大多对二维扫描图使用互相关算法,以数据集中每一帧二维图像为模板图像,其下一帧图像与其做互相关来对齐。现有方法对于黄斑区这种血管特征均匀的组织效果较好,但对于血管特征较复杂的眼底组织结构区域,尤其人眼视盘区的生理结构比黄斑区更为复杂,眼底血管更加丰富,采用常规的黄斑区图像运动伪差的校正方法效果不理想。 本论文依据人眼视盘区OCT图像的结构特点,针对OCT系统的X方向和Y方向的运动伪差,设计了一套针对数据错位、数据重复和数据缺失畸变的体数据的校正算法。通过对C扫描图的像素行数据使用线性相关匹配算法来校正X方向运动伪差,并且利用该方法判断具有运动伪差的图像序列数据断层位置,再判断断层是部分数据重复或者数据缺失。若为数据重复,则利用像素行相关匹配算法来确定数据重复的位置和数目;若为数据断层则需要借助Y方向参考二维扫描图像,来确定数据缺失的位置和数目。然后通过差值平均法对体数据Z方向伪差进行校正,最后对模型体数据和校正后的体数据分别三维重建。利用此方法进行了模型验证,结果显示该方法能够有效还原真实的图像序列,使眼底视网膜的生理结构得到真实的反映。 综上所述:本论文利用提出的基于C扫描图的视盘OCT图像运动伪差校正算法校正伪差后,能够更加真实地反应眼底视盘区的生理结构,进而减少眼科诊断的误诊、漏诊,同时为增加眼科早期诊断的科学性和准确性提供了思路。 |
论文文摘(外文): |
Optical coherence tomography (OCT) is a novel lossless optical image technology, which can be utilized to realize 2D imaging and 3D visualization of morphologies and structures in tissues. OCT has the advantages of high sensitivity and high resolution, and it can be used to implement non-contacting real-time vivo imaging. Besides, the transverse resolution and the longitudinal resolution of OCT are independent. OCT makes enormous contributions to ophthalmology and has become a standard instrument in early diagnosis of eye diseases for its ability to obtain images of retina sections that traditional ophthalmic noninvasive diagnosis cannot acquire. During the dynamic 3D imaging implemented by OCT system, the gathered volume data will generate dislocations and distortions because of the involuntary motion of eyeball, which results in unreality of volume dataset and misdiagnosis in clinical diagnosis. To solve this problem, methods aiming to correct distortions in X and Z directions have been proposed, in which cross-correlation algorithm is used to align every adjacent two images. However, the existing correction methods perform well in tissues like macular region where vascularity is evenly distributed while they are incapable to areas with complicated vessels. And on account of its complex physiological structures and abundant vessels, the optic disc images cannot be corrected appropriately by conventional image rectification methods. In this paper, a correction algorithm aiming at dislocations, repetitions and deficiencies of volume data is proposed to rectify motion artifacts of OCT system. The correction algorithm directs against characteristics of optic disc OCT images. During the correcting course, linear correlation matching algorithm, which can be used to correct motion artifacts of X direction and seek out faults of image sequences, is firstly utilized to process pixel row data of C-scan images. Then, row correlation matching algorithm will be used to ascertain the number and locations of repeated data, if any. And if there are missing data, a reference Y direction scan image is needed to estimate the number and locations. Model validations indicate that the proposed correction algorithm can effectively restore the gathered image sequences and present the real physiologic structures of retina. In summary, the proposed correction algorithm realizes good effects in correcting image data and the corrected images reflect the real physiologic structures of retina. With this correction algorithm, accuracy and scientificity of clinical diagnosis will be enhanced, which leads to effective reduction of misdiagnosis. |
开放日期: | 2016-05-25 |