Ocular OCT Image Quality Assessment For Computer Aided Diagnosis Of Glaucoma

Glaucoma is one of the three worldwide eye diseases. If the eyes are damaged irreversibly, it willeventually lead to blindness. But if this disease can be detected early and treated properly, most patientsstill maintain visual function in a lifetime. Therefore, glaucoma’s early detection, early diagnosis andearly treatment are very important. Optical coherence tomography (OCT) is a new imaging mean in thefield of ophthalmology with non-contact, high-resolution, fast-imaging and other characteristics.However, on the one hand, the OCT images are inevitably affected by speckle noise due to coherentimaging. The decline of the sharpness of an image will have some impact on the diagnostic accuracyand efficiency. On the other hand, the eye OCT image is a massive data. If the selection of clear imagesis based on doctors’ clinical experience, the assessment process will be obviously subjective andinefficiencies. So selecting clear and effective OCT images through computer automatically can reducethe rate of misdiagnosis and doctors’ working strength significantly. This paper researches on theassessment of eye OCT image based on the background.For existing image quality evaluations are not consistent with people’s subjective feeling, a newassessment of anterior chamber OCT image based on the human visual system (HVS) is proposed inthis paper. The fractal dimension is introduced because traditional image quality assessments are singleand usually ignore image structure. And the calculation of this index is improved. On this basis,considering the lack of the assessment by a single fractal dimension, the complexity is introduced forcompensation. The main work in this paper is divided into the following sections:(1) An anterior chamber OCT image assessment based on the HVS is proposed. The traditionalquality assessments including contrast, ambiguity and signal to noise ratio (SNR) are combinedwith HVS characteristics, like visual interest, gray sensitive, contrast sensitivity function and othercharacteristics. Simultaneously, subjective evaluation criterion is drawn up by a doctor and imagescores are given. Finally linear fitting is done between the three indexes and the subjective scores.The experimental results show that the method based on HVS characteristics is more consistentwith the subjective scores than traditional means.(2) Considering the traditional quality assessment is too single and ignore the image structureinformation, the fractal dimension is introduced. It evaluates the images from the similaritybetween the local and global. The traditional calculation of fractal dimension only considers the gray extremes of image pixel and ignores the intermediate gray level. It is improved because of itsdisadvantage. The results show that the improved method has certain enhancement in ordinaryimages. Then the improved method is applied to the retinal OCT image. The correspondingsubjective standard is created and image scores are given by a doctor. The results show that theimproved mean is better than before when used in the OCT image.(3) Considering the fractal dimension can’t reflect the image texture feature as a local statisticalcharacteristic. While the complexity measure thinks that the real mixed signal is between regularsignal and random signal. The similar texture is very important for image quality. Consequently,the complexity measure is introduced based on the single fractal dimension assessment. PCAdimensionality reduction is used for establishing a double parameter model. The experimentalresults indicate that the raised quality assessment is not only superior to traditional methods, butalso better than the single improved fractal dimension. It is the most consistent with the subjectivescores.