Research On Some Problems And Application Of Dual-Wavelength Retinal Oximetry Based On Fundus Camera

A dual-wavelength retinal oximetry based on commercial fundus camera can obtain retinal structural and functional imaging,which provides a useful tool for research of life sciences and diagnosis of retinal-related diseases.However,there exist several problems for the retinal oximetry,such as misalignment between the dual-wavelength images,image noise,and so on.These factors dramatically affect the calculation accuracy and speed of retinal oxygen saturation,and limit clinical application of retinal oximetry.In order to solve above problems,a dual-wavelength retinal oximetry based on fundus camera is designed,including image acquisition system and blood oxygen saturation?SO₂?calculation software system:A new image registration algorithm based on vessel segmentation and optic disc localization is proposed by improving the traditional image registration algorithm.Based on the root cause analysis for misalignment between the dual-wavelength images,dual wavelength registration method based on camera calibration is proposed.Through image noise analysis and parameters estimation,an improved denoising algorithm is proposed.And the denoising algorithm is evaluated from two aspects,including image quality and SO₂calculation accuracy.Finally retinal SO₂ measurement is completed.The specific works mentioned above are summarized as follows:?1?In order to measure retinal SO₂,a dual-wavelength retinal oximetry based on commercial fundus camera is designed.The system makes the retinal image of the commercial fundus camera second imaging,and simultaneously collects dual-wavelength images that respectively represent the same absorption coefficient wavelength images and the different one of oxygenated hemoglobin and de-oxygenated hemoglobin.The retinal SO₂ is finally measured after image registration,vessel segmentation and calculation of optic density ratio of the two images.?2?Because there exists misalignment between the dual-wavelength retinal images,which should be adjusted before they are used to analyze retinal SO₂.This paper presents a new approach to dual-wavelength retinal image registration based on vessel segmentation and optic disc detection.Firstly,the multi-scale segmentation algorithm based on Hessian matrix is used to realize vessel segmentation.Secondly,after accurate optic disc is detected by convergence index filter and the center of the optic disc is obtained by centriod algorithm,the translational difference between the images can be determined.The center of the optic disc is used as the center of rotation,and the registration based on mutual information can be achieved using contour and gray information of vessels through the segmented images.So the rotational difference between the images can be determined too,and the image registration has been realized finally.?3?According to the characteristics of dual wavelength imaging,we first analyze the root reason for misalignment between dual-wavelength retinal images,and conclude that there only exists the rotation and translation deviation between the two images caused by a mounting deviation of CCD cameras.The registration parameters were only related to system itself.Therefore,the corresponding parameters are fixed under the condition of no external force.Based on this,we propose a registration method based on camera calibration.HALCON,which is a classic Machine Vision Software,is used to realize calibration after feature extraction and affine transformation matrix calculation.The method which is not limited to retinal image quality and the accuracy of blood vessel segmentation,can quickly extract feature,and achieve higher precision registration.It has a very important engineering significance.?4?In order to improve image quality and SO₂ calculation accuracy,noise analysis and corresponding denoising algorithm for dual-wavelength images are studied.Firstly,noise parameters are estimated by mixed Poisson-Gaussian?MPG?noise model.Secondly,a novel MPG denoising algorithm which we called VST+DDID is proposed based on variance stabilizing transform?VST?and improved dual domain filter.The results show that VST+DDID is able to effectively remove MPG noise and preserve image edge details.VST+DDID is better than VST+Block-Matching and 3-D filtering?BM3D?,especially in preserving low-contrast details.The following simulation and analysis indicate that MPG noise in the retinal images can lead to erroneously low measurement for SO₂,and the denoised images can provide more accurate grayscale values for retinal oximetry.?5?According to the characteristics of dual-wavelength retinal images,it is found that they satisfy the characteristics of multimode images.In order to maximize the use of complementary information at the edges of dual-wavelength images and further reduce the calculation error of SO₂,we improve VST+DDID.Firstly,noise parameters are also estimated by MPG noise model.Secondly,a novel MPG denoising algorithm which we called VST+CDDID is proposed based on VST and cross dual domain filter.To evaluate the proposed algorithm,both simulative and real experiments have been carried out and the results show that the proposed method can effectively remove MPG noise and preserve edge details.Compared with VST+DDID,the proposed method shows great advantage in terms of PSNR,SSIM and visual quality.The following simulation and analysis indicate that the images denoised by VST+CDDID can provide more accurate grayscale values than the images denoised by VST+DDID for retinal oximetry.?6?Based on the dual-wavelength retinal oximetry,clinical application has been conducted to measure retinal SO₂ in diabetic patients.The work mainly includes comparing the difference of retinal SO₂ between diabetic patients and healthy subjects,and evaluating the repeatability and reliability of the system.The results show that the retinal average arteriolar and venous SO₂ in diabetes group are higher than that in healthy group.The means oxygen saturation of artery and venules in health human for repeated measurements are 96.52%and 57.86%,respectively.The mean standard deviation is 3.62%for arterioles and 4.58%for venules.The means oxygen saturation of artery and venules in diabetic patients for repeated measurements are 100.57%and62.49%,respectively.The mean standard deviation is 4.92%for arterioles and 5.98%for venules.These demonstrate that the system is reliable and reproducible.