Registration of multi-temporal remote sensing images has been widely applied in military and civilian fields, such as ground target identification, urban development assessment, geographic change assessment. Ground surface challenges feature point detection amount quality, which is a common dilemma faced by feature-based registration algorithms. Under severe appearance variation, detected points may contain large proportion outliers, whereas inliers be inadequate unevenly distributed. This...

Corona Virus Disease (COVID-19) has spread globally quickly, and resulted in a large number of causalities medical resources insufficiency many countries. Reverse-transcriptase polymerase chain reaction (RT-PCR) testing is adopted as biopsy tool for confirmation virus infection. However, its accuracy low 60-70%, which inefficient to uncover the infected. In comparison, chest CT been considered prior choice diagnosis monitoring progress COVID-19 Although diagnostic systems based on artificial...

Nasopharyngeal carcinoma (NPC) is a malignant tumor whose survivability greatly improved if early diagnosis and timely treatment are provided. Accurate segmentation of both the primary NPC tumors metastatic lymph nodes (MLNs) crucial for patient staging radiotherapy scheduling. However, existing studies mainly focus on tumors, eliding recognition MLNs, thus fail to comprehensively provide landscape identification. There three main challenges in segmenting MLNs: variable location, size,...

Abstract Breast carcinoma is the second largest cancer in world among women. Early detection of breast has been shown to increase survival rate, thereby significantly increasing patients’ lifespan. Mammography, a noninvasive imaging tool with low cost, widely used diagnose disease at an early stage due its high sensitivity. Although some public mammography datasets are useful, there still lack open access that expand beyond white population as well missing biopsy confirmation or unknown...

Malaria is a significant public health concern, with ∼95% of cases occurring in Africa, but accurate and timely diagnosis problematic remote low-income areas. Here, we developed an artificial intelligence-based object detection system for malaria (AIDMAN). In this system, the YOLOv5 model used to detect cells thin blood smear. An attentional aligner (AAM) then applied cellular classification that consists multi-scale features, local context aligner, attention. Finally, convolutional neural...

Graphs are an effective data structure for characterizing ubiquitous connections as well evolving behaviors that emerge in inter-wined systems. Limited by the stereotype of node-to-node connections, learning node representations is often confined a graph diffusion process where local information has been excessively aggregated, random walk neural networks (GNN) explores far-reaching neighborhoods layer-by-layer. In this regard, tremendous efforts have made to alleviate feature over-smoothing...

Measurement of the width fetal lateral ventricles (LVs) in prenatal ultrasound (US) images is essential for antenatal neuronographic assessment. However, manual measurement LV highly subjective and relies on clinical experience scanners. To deal with this challenge, we propose a computer-aided detection framework automatic LVs two-dimensional US images. First, train deep convolutional network 2,400 to perform pixel-wise segmentation. Then, number pixels per centimeter (PPC), vital parameter...

Land use and land cover (LULC) change is frequent in mountainous terrain of southern China. Although remote sensing technology has become an important tool for gathering monitoring LULC dynamics, image pairs can occur scale changes, noises, geometrical distortions, illuminated variations if these are acquired from different types sensors (e.g., satellites). Meanwhile, how to design efficient detection algorithm that ensures a high rate remains critical challenging step. To address problems,...

COVID-19 causes burdens to the ICU. Evidence-based planning and optimal allocation of scarce ICU resources is urgently needed but remains unaddressed. This study aims identify variables test accuracy predict need for admission, death despite care, among survivors, length stay, before patients were admitted Retrospective data from 733 in-patients confirmed with COVD-19 in Wuhan, China, as March 18, 2020. Demographic, clinical laboratory collected analyzed using machine learning build...

Multiview clustering seeks to partition objects via leveraging cross-view relations provide a comprehensive description of the same objects. Most existing methods assume that different views are linear transformable or merely sampling from common latent space. Such rigid assumptions betray reality, thus leading unsatisfactory performance. To tackle issue, we propose learn both and specific spaces for each view fully exploit their collaborative representations. The space corresponds universal...

Unsupervised feature selection is a vital yet challenging topic for effective data learning. Recently, 2-D methods show good performance on image analysis by utilizing the structure information of image. Current usually adopt sparse regularization to spotlight key features. However, such scheme introduces additional hyperparameter needed pruning, limiting applicability unsupervised algorithms. To overcome these challenges, we design filter estimate weight features selection. Theoretical...

Functional connectivities (FC) of brain network manifest remarkable geometric patterns, which is the gateway to understanding dynamics. In this work, we present a novel geometric-attention neural characterize time-evolving state change from functional neuroimages by tracking trajectory dynamics on high-dimension Riemannian manifold symmetric positive definite (SPD) matrices. Specifically, put spotlight learning common state-specific signatures that represent underlying cognition. context,...

Land degradation, soil erosion and illegal occupation in mountainous terrain of southern China have led to an ever-decreasing stock cultivated land. Small unmanned aerial vehicles (UAVs) are used collect images with very fine spatial temporal resolutions. However, acquired image pairs the same scene often contain scale changes, noises rotated changes at different scales. To address these problems, we propose a small UAV-based multi-temporal change detection for land cover which contains...

Abstract Functional neural activities manifest geometric patterns, as evidenced by the evolving network topology of functional connectivities (FC) even in resting state. In this work, we propose a novel manifold‐based for brain networks (called “Geo‐Net4Net” short) to learn intrinsic low‐dimensional feature representations resting‐state on Riemannian manifold. This tool allows us answer scientific question how spontaneous fluctuation FC supports behavior and cognition. We deploy set positive...

Land degradation, soil erosion, and illegal occupation in the mountainous terrain of southern China have severely reduced amount cultivatable land. The use small unmanned aerial vehicles (UAVs, aka drones) equipped with various types cameras is considered to be a flexible low-cost platform for monitoring cultivated land changes. However, image pairs same scene taken from different viewpoints often contain discontinuous rotated images illuminated variations. To address these problems, novel...

Fundus image is commonly used in aiding the diagnosis of ophthalmic diseases. A high-resolution (HR) valuable to provide anatomic information on eye conditions. Recently, super-resolution (SR) though learning model has been shown be an economic yet effective way satisfy high demands clinical practice. However, reported methods ignore mutual dependencies low-and images and did not fully exploit between channels. To tackle with drawbacks, we propose a novel network for fundus SR, named by...

Accurate nasopharyngeal carcinoma (NPC) segmentation in magnetic resonance image (MRI) is crucial for diagnosis and treatment. However, most existing deep learning methods performed unsatisfactorily, since NPC infiltrative typically has a small or even tiny volume with indistinguishable boundary, making it indiscernible from tightly connected surrounding tissue immense complex background. To address the background dominant problem, this paper proposes coarse-to-fine model. The proposed model...

Neural ordinary differential equation (ODE) can be approximated by nonlinear mappings using continuous-times ODEs such that a family of models is formed. Due to their desirable properties, as invertibility, stability, robustness and parameter efficiency, neural have attracted increasing attention recently. However, its performance (e.g., stability robustness) on neuroimaging data has not been explored. To fill this gap, we propose an ODE-based brain state recognition network (OSR-Net). It...

Nonrigid point set registration is a key technology in the field of remote sensing image registration, which widely used military and civil fields such as natural disaster damage assessment, agricultural urban land-use planning, environmental quality monitoring, ground target identification. We present multifeature energy optimization framework parameter adjustment-based nonrigid that has three contributions: (1) an designed to freely combine multiple features for estimating correspondences...