Research on Fundus Image Mosaic Method Based on Genetic Algorithm

Retinal image mosaic is the key to detect common diseases, and the existing image mosaic methods are difficult to solve the problems of low contrast of fundus images and geometric distortion between images in different fields of view. To solve the problem of noise in retinal fundus images, an image mosaic algorithm based on the genetic algorithm was proposed. Firstly, a series of morphological pretreatment was performed on the fundus images. Then, the vascular network is extracted by obtaining the maximum entropy of the image to determine the threshold value. The similarity of the image to be spliced is a feature, and the genetic algorithm is used to solve the optimal parameters to achieve the maximum similarity. By smoothing the image, a clear image with minimum noise is obtained. Experimental results show that the proposed algorithm can effectively realize the image mosaic of the fundus. The method proposed in this paper can provide support for high-precision automatic stitching of multiple single-mode color fundus images.

Development of deep learning-based detecting systems for pathologic myopia using retinal fundus images

Globally, cases of myopia have reached epidemic levels. High myopia and pathological myopia (PM) are the leading cause of visual impairment and blindness in China, demanding a large volume of myopia screening tasks to control the rapid growing myopic prevalence. It is desirable to develop the automatically intelligent system to facilitate these time- and labor- consuming tasks. In this study, we designed a series of deep learning systems to detect PM and myopic macular lesions according to a recent international photographic classification system (META-PM) classification based on color fundus images. Notably, our systems recorded robust performance both in the test and external validation dataset. The performance was comparable to the general ophthalmologist and retinal specialist. With the extensive adoption of this technology, effective mass screening for myopic population will become feasible on a national scale.

Deep Multi-Segmentation Approach for the Joint Classification and Segmentation of the Retinal Arterial and Venous Trees in Color Fundus Images

The analysis of the retinal vasculature represents a crucial stage in the diagnosis of several diseases. An exhaustive analysis involves segmenting the retinal vessels and classifying them into veins and arteries. In this work, we present an accurate approach, based on deep neural networks, for the joint segmentation and classification of the retinal veins and arteries from color fundus images. The presented approach decomposes this joint task into three related subtasks: the segmentation of arteries, veins and the whole vascular tree. The experiments performed show that our method achieves competitive results in the discrimination of arteries and veins, while clearly enhancing the segmentation of the different structures. Moreover, unlike other approaches, our method allows for the straightforward detection of vessel crossings, and preserves the continuity of the arterial and venous vascular trees at these locations.

Latent Diabetic Macular Edema in Chinese Diabetic Retinopathy Patients

Purpose: To compare the detection rates of optical coherence tomography (OCT) and fluorescein angiography (FA) in a diabetic macular edema (DME) and the severity of diabetic retinopathy in both color fundus images (CFI) and FA, and to investigate the predictive factors in macular leakages in FA.Methods: This was a retrospective study, and a total of 132 eyes of 77 patients with diabetic retinopathy were enrolled. Macular OCT, FA, and CFI were reviewed and measured. Central foveal thickness was also measured.Results: The severity of diabetic retinopathy in FA was significantly higher than that in CFI (p < 0.001). OCT detected 26 eyes with DMEs, which included the following: 13 eyes with cystoid macular edemas; 13 eyes with serous retinal detachments; 11 eyes with diffuse retinal thickening; 4 eyes with vitreomacular interface abnormalities. In contrast, 72 out of 132 eyes (54.5%) showed macular leakages in FA, which was significantly higher than that detected by OCT (p < 0.001). Compared with FA, the sensitivity and the specificity of OCT in detecting DMEs were 30.6 and 93.3%, respectively. However, central foveal thickness was not significantly different between the patients with non-clinically significant macular edema (CSME, 253.1 ± 26.95 μm) and slight CSME (270.9 ± 37.11 μm, p = 0.204). The mean central foveal thickness in diabetic macular edema (FA) eyes was 271.8 ± 66.02 μm, which was significantly higher than that (253. ± 25.21 μm) in non-DME (FA) eyes (p = 0.039). The central foveal thickness in DME (FA) eyes was significantly lower than that in eyes with DME (OCT) (p = 0.014). After adjusting for age and sex, a logistic regression analysis showed that the classification of diabetic retinopathy in FA was positively associated with macular leakage in FA (p < 0.001).Conclusions: The severity of diabetic retinopathy is underestimated in CFI compared with that in FA. FA can detect latent DMEs, which appeared normal on OCT. The central foveal thickness is not a sensitive parameter for detecting latent DMEs.

Diabetic Retinopathy Detection using Retinal Images and Deep Learning Model

Diabetic Retinopathy (DR) is one of the serious problems caused by diabetes and a leading source of blindness in the working-age population of the advanced world. Detecting DR in the early stages is crucial since the disease generally shows few symptoms until it is too late to provide an effective cure. But detecting DR requires a skilled clinician to examine and assess digital color fundus images of the retina. By simplifying the detection process, severe damages to the eyes can be prevented. Many deep learning models particularly Convolutional Neural Networks (CNNs) have been tested in similar fields as well as in the detection of DR in early stages. In this paper, we propose an automatic model for detecting and suggesting different stages of DR. The work has been carried out on APTOS 2019 Blindness Detection Benchmark Dataset which contains around 3600 retinal images graded by clinicians for the severity of diabetic retinopathy on a range of 0 to 4. The proposed method uses ResNet50 (Residual Network that is 50 layers deep) CNN model along with pre-trained weights as the base neural network model. Due to its depth and better transfer learning capabilities, the proposed model with ResNet50 achieved 82% classification accuracy. The classification ability of the model was further analysed with Cohen Kappa score. The optimized validation Cohen Kappa score of 0.827 indicate that the proposed model didn’t predict the outputs by chance.

Using Artificial Intelligence Reading Label System in Diabetic Retinopathy Grading Training of Junior Ophthalmology Residents and Medical Students

Purpose Evaluate the efficiency of using artificial intelligence reading label system in diabetic retinopathy grading training of junior ophthalmology resident doctors and medical students. Methods Loading 520 diabetic retinopathy patients’ color fundus images in the artificial intelligence reading label system. 13 participants (including 6 junior ophthalmology residents and 7 medical students) read the images randomly for 8 rounds. They evaluated the grading of images and labeled the typical lesions. The sensitivity, specificity and kappa score were determined by comparison with the participants’ results and expert golden standards. Results Through 8 round reading, average kappa score was elevated from 0.67 to 0.81. Average kappa score of round 1 to 4 was 0.77, and average kappa score of round 5 to 8 was 0.81. The participant was divided into two groups. Participants in group 1 were junior ophthalmology resident students and participants in group 2 were medical doctors. Average kappa score of group 1 was elevated from 0.71 to 0.76. Average kappa score of group 2 was elevated from 0.63 to 0.84. Conclusion The artificial intelligence reading label system was a useful tool in training resident doctors and medical students in doing diabetic retinopathy grading.

Multi-label classification of fundus images based on graph convolutional network

Background Diabetic Retinopathy (DR) is the most common and serious microvascular complication in the diabetic population. Using computer-aided diagnosis from the fundus images has become a method of detecting retinal diseases, but the detection of multiple lesions is still a difficult point in current research. Methods This study proposed a multi-label classification method based on the graph convolutional network (GCN), so as to detect 8 types of fundus lesions in color fundus images. We collected 7459 fundus images (1887 left eyes, 1966 right eyes) from 2282 patients (1283 women, 999 men), and labeled 8 types of lesions, laser scars, drusen, cup disc ratio ($$C/D>0.6$$ C / D > 0.6 ), hemorrhages, retinal arteriosclerosis, microaneurysms, hard exudates and soft exudates. We constructed a specialized corpus of the related fundus lesions. A multi-label classification algorithm for fundus images was proposed based on the corpus, and the collected data were trained. Results The average overall F1 Score (OF1) and the average per-class F1 Score (CF1) of the model were 0.808 and 0.792 respectively. The area under the ROC curve (AUC) of our proposed model reached 0.986, 0.954, 0.946, 0.957, 0.952, 0.889, 0.937 and 0.926 for detecting laser scars, drusen, cup disc ratio, hemorrhages, retinal arteriosclerosis, microaneurysms, hard exudates and soft exudates, respectively. Conclusions Our results demonstrated that our proposed model can detect a variety of lesions in the color images of the fundus, which lays a foundation for assisting doctors in diagnosis and makes it possible to carry out rapid and efficient large-scale screening of fundus lesions.