A Teleophthalmology Support System Based on the Visibility of Retinal Elements Using the CNNs.

Rapid development in the field of ophthalmology has increased the demand of computer aided diagnosis of various eye diseases. Papilledema is an eye disease in which the optic disc of the eye is swelled due to an increase in intracranial pressure. This increased pressure can cause severe encephalic complications like abscess, tumors, meningitis or encephalitis, which may lead to a patient's death. Although there have been several papilledema case studies reported from a medical point of view, only a few researchers have presented automated algorithms for this problem. This paper presents a novel computer aided system which aims to automatically detect papilledema from fundus images. Firstly, the fundus images are preprocessed by going through optic disc detection and vessel segmentation. After preprocessing, a total of 26 different features are extracted to capture possible changes in the optic disc due to papilledema. These features are further divided into four categories based upon their color, textural, vascular and disc margin obscuration properties. The best features are then selected and combined to form a feature matrix that is used to distinguish between normal images and images with papilledema using the supervised support vector machine (SVM) classifier. The proposed method is tested on 160 fundus images obtained from two different data sets i.e. structured analysis of retina (STARE), which is a publicly available data set, and our local data set that has been acquired from the Armed Forces Institute of Ophthalmology (AFIO). The STARE data set contained 90 and our local data set contained 70 fundus images respectively. These annotations have been performed with the help of two ophthalmologists. We report detection accuracies of 95.6% for STARE, 87.4% for the local data set, and 85.9% for the combined STARE and local data sets. The proposed system is fast and robust in detecting papilledema from fundus images with promising results. This will aid physicians in clinical assessment of fundus images. It will not take away the role of physicians, but will rather help them in the time consuming process of screening fundus images.

Automatic detection of optic disc and fovea is a precursor to the computer-aided analysis of retinal pathologies. In this paper, we present a unified approach for optic disc and fovea detection based on normalized cross-correlation technique. The algorithm performance is optimized by introducing vector inner products and norms instead of conventional mean and variance computations. We report optic disc detection results on four publicly available fundus image databases amounting to a total of 1451 fundus images and fovea detection results on another four publicly available fundus image databases amounting to a total of 1454 fundus images. The proposed method results in an optic disc detection accuracy of 99.01%, 95.67%, 99.09%, and 100% on DRISHTI-GS, MESSIDOR, DRIONS-DB, and DRIVE fundus image databases, respectively, and fovea detection accuracy of 94.83%, 84.62%, 95.51%, and 97.14% on MESSIDOR, DIARETDB0, DIARETDB1, and DRIVE fundus image databases, respectively. The speed of optic disc and fovea detection has been improved considerably by downsampling technique. In addition, we report the effect of downsampling on the detection accuracy.

Efficient Optic Disc(OD) detection and segmentation is an essential element in the design of any automated retinal image analysis system. The detection of OD is often challenging due to the uneven illumination and lack of a definite OD boundary which is further complicated in infant retinal images due to their low quality. In this paper, a method for efficient localization and segmentation of OD in infant fundus images is proposed utilizing the commonly occurring double ring structure in the OD of premature infants. A method to assign an OD score to each pixel position is proposed which corresponds to the likelihood that the pixel belongs to the OD area, thus highlighting the possible locations of OD. The real location of OD is finalized after post-processing, and then the OD shape is segmented using a single seed region growing. The proposed method is extremely robust to the variations in the quality of the infant fundus images with 98.1% accuracy in locating the OD when tested on 317 infant fundus images of varying quality.

Rapid, bloody, and blinding diabetic retinopathy

Evaluation of a deep learning system for automatic detection of proximal surface dental caries on bitewing radiographs

Aim: To evaluate a new algorithm to detect the optic disc in retinal fundus images on a number of publicly available datasets. Optic Disc detection is an important first step in many automated algorithms, either to be masked out of future processing or for the use in optic disc related disease such as glaucoma and papilledema.Methods: We propose a new method for optic disc detection that converts the retinal image into a graph and exploits vessel enhancement methods to calculate edge weights in finding the shortest path between pairs of points on the periphery of the image. The line segment with the maximum number of shortest paths is considered the optic disc location, with refinement from a combination template matching approach in the found region. The method was tested on three publicly available datasets: DRIVE, DIARETDB1, and Messidor consisting of 40, 89, and 1200 images. All images were acquired at a 45°-50° field of view.Results: The method achieves an accuracy of 100, 98.88, and 99.42% on the DRIVE, DIARETDB1, and Messidor databases respectively.Conclusions: The method performs as well or better than state-of-the-art methods on these datasets. Processing takes an average of 32 seconds (+-1.2) to detect the optic disc, with 26 of those seconds used for the vessel enhancement process. The accuracy over a wide variety of images shows that the method is robust and would be optimal for retinal analysis systems that perform vessel enhancement as part of their processing.

Knowledge of the number and distribution of oil palm trees during the crop cycle is vital for sustainable management and predicting yields. The accuracy of the conventional image processing method is limited for the hand-crafted feature extraction method and the overfitting problem occurs due to the insufficient dataset. We propose a modification of the Faster Region-based Convolutional Neural Network (FRCNN) for palm tree detection to reduce the overfitting problem and improve the detection accuracy. The enhanced FRCNN (EFRCNN) leads to improved performance for detecting objects (in the same image) when they are of multiple sizes by using a feature concatenation method. Transfer learning based on a ResNet50 model is used to extract the features of the input image. High-resolution images of oil palm trees from a drone are used to form the data set, containing mature, young, and mixed oil palm tree regions. We train and test the EFRCNN, the FRCNN, a CNN used recently for oil palm image detection, and two standard methods, namely, the support vector machine (SVM) and template matching (TM). The results reveal an overall accuracy of ≥96.8% for the EFRCNN on the three test sets. The accuracy is higher than the CNN and FRCNN and substantially higher than SVM and TM. For large-scale plantations, the accuracy improvement is significant. This research provides a method for automatically counting the oil palm trees in large-scale plantations.

Cup-to-disc ratio is an important measure to diagnose glaucoma. This measure can be automatically computed from the segmentation of the optic disc and cup in eye-fundus image. In this paper, a novel segmentation algorithm based on Soft Computing techniques is presented. This theory is able to handle the imprecision and uncertainty present in the determination of the boundaries of these structures. The algorithm is composed of three main steps: vessel segmentation and removal, inpainting and optic disc and cup boundaries localisation. The preliminary results show the potential of this approach which obtains a visual accurate segmentation of both structures in the images of the DRIVE database.

Optic disc detection and segmentation using saliency mask in retinal fundus images

Regular eye examinations are the key to limiting the vision loss caused by glaucoma and diabetic retinopathy. Optic disc (OD) detection is of vital importance in developing automated diagnosis systems for these diseases. In this work we present a method for automatic localization and boundary detection of the optic disc in retinal fundus images. In the first step, we rely on the OD geometric feature and utilize gaussian and mean curvatures for the localization of the OD. In the second step, we extract a region of interest based on the OD localization. Then, to the edges of this extracted region a circular Hough transform is applied to segment the OD boundary. The experimental results on three public datasets show the efficacy of the proposed method.

WITHDRAWN: Effective utilization of YOLOv3 model for aircraft detection in Remotely Sensed Images

PurposeThe retina could serve as a window of neuroinflammation, but the in vivo changes in macrophage-like cell (MLC), such as microglia, in acute ischemic retinal stroke remain unclear. Thus, the current study aimed to investigate the in vivo changes in MLC characterized by en face optical coherence tomography (OCT) after acute ischemic retinal stroke.MethodsTwenty patients with unilateral acute nonarteritic reperfused central retinal artery occlusion (CRAO) were participated in this study, and their contralateral eyes served as control group. A 3 μm en face OCT slab on the inner limiting membrane of the optic nerve head (ONH) region or macular region was used to visualize and binarize the MLCs. The MLCs were binarized and quantified using a semiautomated method. OCT angiography was used to evaluate the reperfusion status and obtain the structural data of the inner retina in the ONH and macula. The thickness of the ganglion cell complex in the macular region was measured. The optical intensity and optical intensity ratio of the inner retina were calculated to evaluate the ischemia severity.ResultsIn the ONH region, decreased vessel densities of radial peripapillary capillaries accompanied by increased thickness of the retinal nerve fiber layer were found in the CRAO eyes in comparison to the unaffected eyes (p=0.001, p=0.009, respectively). In the macular region, significantly lower vessel densities in both the superficial and deep capillary plexus and increased thickness of the ganglion cell complex were also found in the CRAO eyes (all p ≤ 0.001). The ONH and macular MLC quantities and densities in CRAO eyes were significantly higher than those in the unaffected eyes (both p<0.001). Larger and plumper MLCs were observed in the CRAO eyes compared with their unaffected eyes. ONH and macular MLC densities were positively associated with the disease duration in the acute phase and the optical intensity ratio of inner retina.ConclusionsThe increased density and morphological changes of MLCs may indicate the aggregation and activation of MLCs following acute reperfused CRAO. The aggregation of MLCs may be more pronounced in CRAO eyes with longer disease duration and more severe ischemia. MLCs characterized by en face OCT may serve as an in vivo visual tool to investigate neuroinflammation in the ischemic-reperfusion process of stroke.

Introduction: This study aims to investigate the changes of retinal vascular system in primary angle closure glaucoma (PACG) and acute primary angle closure (APAC) by optical coherence tomography (OCT) angiograph (OCTA) and to evaluate the diagnostic ability of changes of vessel density (VD) in different sectors and layers of optic disc and macular area in APAC and PACG. Methods: In this cross-sectional, observational study, 21 APAC patients (22 eyes) and 21 PACG patients (27 eyes) along with 17 healthy people were enrolled from August 2018 to March 2019. Optic disc region and macular region were imaged using swept-source OCTA system. VD of the macular region was quantified by Image J (1.52a, USA) and Matlab 2018a. The circumpapillary retinal nerve fiber layer (cpRNFL) thickness and ganglion cell complex thickness were obtained by spectral-domain OCT. Results: Compared with the healthy group, the cpRNFL thickness in superior sector was thicker in the APAC group, and this area had the most diffuse microvascular dropout as well. The difference in the macular superficial capillary plexus (SCP) VD between APAC and the control group was not statistically significant. The area under the ROC curves (AUC) of the total optic disc VD in the radial peripapillary capillary (RPC) layer was higher than the AUC of the papillary VD in the optic nerve head (ONH) layer. Compared to the control group, the total optic disc VD, peripapillary VD, and each quadrant of peripapillary VD were decreased in PACG (p < 0.01). In PACG macular region, SCP VD, and deep capillary plexus (DCP) VD, parafovea VD (except temporal sectors) decreased (p < 0.01). The PACG eyes had a greater decrease percentage of VD in total ONH than total macula. The diagnostic value of the VD in the ONH layer and the RPC layer was similar. The diagnostic value of the SCP VD in the macula was greater than the DCP VD in the macula. The AUC was no significant difference between cpRNFL thickness and the total optic disc VD AUC. Conclusion: Elevated intraocular pressure preferentially affects vascular perfusion in the optic disc region more than the macular region in APAC and PACG. In the APAC eyes, there was a perfusion defect in the optic disc region and an increase in RNFL thickness. In this study, the OCTA vascular parameters have similar performance to the OCT structural parameters for glaucoma diagnosis in PACG.

Yuan, H.; Zhang, S.; Chen, G., and Yang, Y., 2020. Underwater image fish recognition technology based on transfer learning and image enhancement. In: Hu, C. and Cai, M. (eds.), Geo-informatics and Oceanography. Journal of Coastal Research, Special Issue No. 105, pp. 124–128. Coconut Creek (Florida), ISSN 0749-0208.To effectively identify fish targets in underwater images, an image processing technology based on secondary migration learning and the Retinex algorithm is proposed to solve the problem of few underwater data sets and unclear underwater images. This method only uses a small-scale underwater image data set to train the network, and the Faster region-based convolutional neural networks (R-CNN) model can quickly detect underwater fish targets. The first transfer learning is applied between the ultra–large-scale ImageNet open-source dataset and the medium-scale Open Images high-definition fish dataset. Then, the Retinex iterative algorithm is used to enhance the underwater image to apply the second transfer learning between the high-resolution medium-scale fish data set and the small-scale underwater data set. Experiments show that this method can train a very effective detection model using small underwater image data sets at a low cost. The effect and performance of the detection model are far superior to traditional machine learning methods. This research can provide a certain reference value for deep-sea exploration, resource protection, and other engineering applications.

Insect detection and control at an early stage are essential to the built environment (human-made physical spaces such as homes, hotels, camps, hospitals, parks, pavement, food industries, etc.) and agriculture fields. Currently, such insect control measures are manual, tedious, unsafe, and time-consuming labor dependent tasks. With the recent advancements in Artificial Intelligence (AI) and the Internet of things (IoT), several maintenance tasks can be automated, which significantly improves productivity and safety. This work proposes a real-time remote insect trap monitoring system and insect detection method using IoT and Deep Learning (DL) frameworks. The remote trap monitoring system framework is constructed using IoT and the Faster RCNN (Region-based Convolutional Neural Networks) Residual neural Networks 50 (ResNet50) unified object detection framework. The Faster RCNN ResNet 50 object detection framework was trained with built environment insects and farm field insect images and deployed in IoT. The proposed system was tested in real-time using four-layer IoT with built environment insects image captured through sticky trap sheets. Further, farm field insects were tested through a separate insect image database. The experimental results proved that the proposed system could automatically identify the built environment insects and farm field insects with an average of 94% accuracy.