• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.401-403
• /
• 2021

본 시스템은 같은 환자로부터 촬영한 기존 안저영상과 초광각 안저영상을 Paired Dataset으로 지니고 있으며, 영상의 크기 및 해상도를 똑같이 맞추고, 황반부와 신경유두 및 혈관의 위치를 미세조정하는 과정을 자동화하는 것을 목표로 하고 있다. 이 과정은 황반부를 중심으로 하여 영상을 잘라내어 이미지의 크기를 맞추는 과정(Scaling)과, 황반부를 중심으로 잘라낸 한 쌍의 영상을 포개었을 때 황반부, 신경 유두, 혈관 등의 위치가 동일하도록 미세조정하는 과정(Warping)이 있다. Scaling Stage에선 기존 안저영상과 초광각 안저영상의 촬영범위가 현저하게 차이나기 때문에, 황반변성 부위를 잘 나타내도록 사전에 잘라낼 필요가 있으며, 이를 신경유두의 Object Detection을 활용할 예정이다. Warping Stage에선 동일한 위치에 같은 황반변성 정보가 내포되어야 하므로 규격조정 및 위치조정 과정이 필수적이며, 이후 안저영상 내의 특징들을 매칭하는 작업을 하기 위해 회전, 회절, 변환 작업 등이 이루어지며, 이는 Homography Estimation을 통하여 이미지 변환 matrix를 구하는 방법으로 진행된다. 자동조정된 안저영상 데이터는 추후에 GAN을 이용한 안저영상 생성모델을 위한 학습데이터로 이용할 예정이며, 현재로선 2500쌍의 데이터를 대상으로 실험을 진행중이지만, 최종적으로 3만 쌍의 안저영상 데이터를 목표로 하고 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.335-337
• /
• 2021

안저 영상 촬영기술이 발달되며 진단에 사용되는 안저 영상에는 시각적으로 많은 변화가 일어났다. 새로운 촬영 기법인 초광각 안저 영상은 기존 영상에 비해 넓은 범위의 영상을 생성할 수 있다. 촬영 범위가 넓어짐에 따라 이미지에는 왜곡이 발생하고, 이로 인해 안저 영상을 통한 황반 부위 진단에 어려움을 야기하기도 한다. 본 논문에서는 이러한 왜곡을 보정하고 초광각 안저 영상을 기존 안저 영상의 영역으로 변환하는 시스템을 강화학습을 통해 구축한다. 제안하는 방법은 A3C 강화학습법을 사용하며 실험 결과는 제안 방법을 통해 안저 영상을 자동으로 변환할 수 있음을 보여준다.

• Journal of Korea Multimedia Society
• /
• v.12 no.8
• /
• pp.1082-1088
• /
• 2009

Fundus images are highly useful in evaluating patients' retinal conditions in diagnosing eye diseases. In particular, vessel regions are essential in diagnosing diabetes and hypertension. In this paper, we used top-hat filter to compensate for non-uniform background. Image contrast was enhanced by using contrast limited adaptive histogram equalization (CLAHE) method. Hessian matrix was next applied to detect vessel regions. Results indicate that our method is 1.3% more accurate than matched filter method. Our proposed method is expected to contribute to diagnosing eye diseases.

• Journal of Korea Multimedia Society
• /
• v.8 no.4
• /
• pp.492-501
• /
• 2005

In this paper, we proposed the new segmentation method of optic nerve head and optic cub to consider the depth of optic nerve head on stereo fundus image. We analyzed the error factor of stereo matching on stereo fundus image, and compensated them. For robust extraction of optic nerve head and optic cub, we proposed the modified active contour model to consider the 3D depth of optic nerve head. As experiment result to various stereo fundus images, we confirmed that proposed method can segment optic nerve head and optic cup effectively.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.636-638
• /
• 2022

This paper is a study to apply CNN(Convolution Neural Network) to fundus images for identifying glaucoma. Fundus images are evaluated in the field of medical diagnosis detection, which are diagnosing of blood vessels and nerve tissues, retina damage, various cardiovascular diseases and dementia. For the experiment, using normal image set and glaucoma image set, two types of image set are classifed by using AlexNet. The result performs that glaucoma with abnormalities are activated and characterized in feature map.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.124-126
• /
• 2021

In this study, fundus images were analyzed using an image processing algorithm for diagnosis of diabetic retinopathy, and specific areas such as hard exudate and retinal hemorrhage, which are characteristic of diabetic retinopathy disease using the bit plane technique, were extracted. We propose a system capable of automatic diagnosis by quantifying the characteristics of diabetic retinopathy based on the analyzed fundus image.

• Journal of Biomedical Engineering Research
• /
• v.24 no.2
• /
• pp.91-97
• /
• 2003

The Propose of this paper is hierarchical detection method for the optic disc in fundus image. We detected the optic disc boundary by using the Prior information. It is based on the anatomical knowledge of fundus which are the vessel information. the image complexity. and etc. The whole method can be divided into three stages . First, we selected the region of interest(ROI) which included optic disc region. This is used to calculate location and size of the optic disc which are prior knowledge to simplify image preprocessing. And then. we divided the fundus image into numberous regions with watershed algorithm and detected intial boundary of the optic disc by reducing the number of the separated regions in ROI. Finally, we have searching the defective parts of boundary as a result of serious vessel interference in order to detect the accurate boundary of optic disc and we have removing and interpolating them.

• Progress in Medical Physics
• /
• v.2 no.2
• /
• pp.183-192
• /
• 1991

Ophthalmoscopy following the intravenous injection of fluorescein has gained great diagnostic importance in ophthalmology. This technique provides sequential evaluation of the anatomic and physiologic status of the choroidal and retinal vasculature. In order to detect the changes between fluorescein ocular fundus image frames, the direct subtraction of the two frames is inadequate because of geometric distortions and background gray level differences in two images. In this study, a scheme for the correction of the geometric distortions is proposed. Precise control point coordinate values for transformation functions are manually determined after the process including a series of blood vessel detection and thinning, and one frame is mapped to another, and then a geometric distortion corrected image is obtained. When the corrected image is used in interframe change detections, a sucessful result is ensured.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.29-37
• /
• 2021

In this paper, we propose a deep learning-based retinal vessel segmentation model for handling multi-scale information of fundus images. we integrate the selective kernel convolution into U-Net-based convolutional neural network. The proposed model extracts and segment features information with various shapes and sizes of retinal blood vessels, which is important information for diagnosing eye-related diseases from fundus images. The proposed model consists of standard convolutions and selective kernel convolutions. While the standard convolutional layer extracts information through the same size kernel size, The selective kernel convolution extracts information from branches with various kernel sizes and combines them by adaptively adjusting them through split-attention. To evaluate the performance of the proposed model, we used the DRIVE and CHASE DB1 datasets and the proposed model showed F1 score of 82.91% and 81.71% on both datasets respectively, confirming that the proposed model is effective in segmenting retinal blood vessels.

• Journal of Korea Multimedia Society
• /
• v.12 no.8
• /
• pp.1073-1081
• /
• 2009

In this paper, we proposed new method to enhance choroidal vessels by suppressing retina vessels brightness. It is well-known that CNV(choroidal neovascularization) is related with sight loss. The main feature of CNV is the occurrence of new vessels in choroid. Unfortunately, because retina vessels brightness is stronger than choroidal vessels brightness in ICGA(indocynanine green angiography) image, so that the choroidal vessels were hardly recognized. Therefore, for correct diagnosis, the choroidal vessels must be enhanced in ICGA image. The proposed enhancement method consists of 3 strategies. First, the retina vessels were detected by multi scale enhancement technique, hysteresis thresholding, KNN(Kth-nearest neighbor) classification method. And then, a retina vessel mask was generated from detection result. Next, the brightness of retina vessels was suppressed by the proposed conditional region erosion method and mask region until the mask region was vanished. Finally, the brightness of choroidal vessel was enhanced on processed image. Through an experiment, we had confirmed that the proposed method was robust and efficient.