Abstract. We propose a convolution neural network based algorithm for simultaneously diagnosing diabetic retinopathy and highlighting suspicious regions. Our contributions are two folds: 1) a network termed Zoom-in-Net which mimics the zoom-in process of a clinician to examine the retinal images. Trained with only image-level supervisions, Zoomin-Net can generate attention maps which highlight suspicious regions, and predicts the disease level accurately based on both the whole image and its high resolution suspicious patches. 2) Only four bounding boxes generated from the automatically learned attention maps are enough to cover 80% of the lesions labeled by an experienced ophthalmologist, which shows good localization ability of the attention maps. By clustering features at high response locations on the attention maps, we discover meaningful clusters which contain potential lesions in diabetic retinopathy. Experiments show that our algorithm outperform the stateof-the-art methods on two datasets, EyePACS and Messidor.
Marine vertical cable seismic (VCS) is a promising survey technique for submarine complex structure imaging and reservoir monitoring, which uses vertical arrays of hydrophones deployed near the seafloor to record seismic wavefields in a quiet environment. Recently, we developed a new type of distributed VCS system for exploration and development of natural gas hydrates preserved in shallow sediments under the seafloor. Using this system and air-gun sources, we accomplished a 3D VCS yield data acquisition for gas hydrates exploration in the Shenhu area, South China Sea. In view of the characteristics of VCS geometry, we implement reverse time migration (RTM) on a common receiver gather to obtain high-resolution images of marine sediments. Due to the unique acquisition method, it is asymmetrical for the reflection path between the sources and the receivers in the VCS survey. Therefore, we apply accurate velocity analysis to common scatter point (CSP) gathers generated from common receiver gathers instead of the conventional velocity analysis based on common depth point gathers. RTM with this reliable velocity model results in high-resolution images of submarine hydrate-bearing sediments in deep water conditions. The RTM imaging section clearly shows the bottom simulating reflector (BSR) and also the reflection characteristics of the hydrate-bearing sediments filled with consolidated hydrates. Moreover, its resolution is relative to that of acoustic logging curves from the nearby borehole, and this imaging section is well consistent with the synthetic seismogram trace generated by the logging data. All these results reveal that VCS is a great potential technology for exploration and production of marine natural gas hydrates.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.