Research On The Improvement Of Retinal Adaptive Optics Imaging Quality

The retinal capillaries are the only microvasculature, which can be noninvasively observed and imaged in vivo. Any systematic and blood pathological lesion will cause the microangiopathy of retina. Therefore, high resolution and high contrast retinal microvacular images make great sense to both the diagnosis of systematic as well as blood diseases and the pathology research. By means of adaptive optics imaging technique, the resolution of retinal image can be effectively improved. However, due to the low accuracy of ocular aberration detection and poor dynamic range of wavefront sensor, the ocular aberration detection of high myopia and astigmatism is unsatisfactory. On the other hand, due to the dynamic behavior of ocular aberrations, there will be residual aberrations during the imaging procedure. It will degrade the resolution and contrast of retinal images. In order to solve this problem, blind deconvolution is always employed to process retinal images. However, the result is always unsatisfactory. Therefore, this paper focuses on the improvement of retinal adaptive optics imaging quality.After a detail study of retinal adaptive optics imaging system and its key components, a systematic analysis of the measurement accuracy and the dynamic range of Shack-Hartmann wavefront sensor are conducted. Basing on this analysis, a novel centroid detection algorithm is introduced. This algorithm gets rid of the restriction of subaperture in classical centroid detection algorithm.Matched filter method is used to detect all of the centroids from the whole spot image at the same time. In contrast to classical centroid detection algorithm, it can improve the measurement accuracy by more than an order of magnitude. After that, a geometry method is used to relate the centroid array with the microlens array. The dynamic range of Shack-Hartmann wavefront sensor is significantly improved. According to the results, the dynamic range can be extended from 57.1% to 160% for the first 24 items of Zernike wavefronts respectively.Due to the dynamic behavior of ocular aberrations, the dynamic ocular aberrations can not be compensated completely during the imaging procedure. The residual ocular aberrations will decrease the quality of retinal images and make it difficult to diagnose diseases. To overcome this obstacle, an image restoration algorithm based on optical transfer function estimation is introduced. This algorithm use Levy Stable distribution to approximate the optical transfer function, which is caused by residual ocular aberrations. Then, a series of retinal images with different restoration results are obtained. The “best” image is selected to diagnose diseases by the image researchers or medical researchers according to the visual effects. The experimental results show that the contrast and resolution of retinal images is significantly improved.Due to the different imaging mechanism between ocular cells and capillaries, ocular cells are white spots with high contrast while ocular capillaries are dark streaks with low contrast in retinal image. Therefore, it is too difficult to segment ocular capillaries from single retinal image with low contrast. To solve this problem, a retinal image segmentation algorithm based on image series analysis is introduced. This algorithm segments ocular capillaries by tracing the motion of blood components. Then, a double-mask matched filter method is also adopted to eliminate the edges, which are caused by the uneven background. Finally, a corner/edge detector is used to eliminate the elongated fragments caused by the vibration of human eyes. The experimental results show that ocular capillaries can be segmented effectively.