Simultaneous Segmentation of Fetal Hearts and Lungs for Medical Ultrasound Images via an Efficient Multi-scale Model Integrated With Attention Mechanism

Large scale early scanning of fetuses via ultrasound imaging is widely used to alleviate the morbidity or mortality caused by congenital anomalies in fetal hearts and lungs. To reduce the intensive cost during manual recognition of organ regions, many automatic segmentation methods have been proposed. However, the existing methods still encounter multi-scale problem at a larger range of receptive fields of organs in images, resolution problem of segmentation mask, and interference problem of task-irrelevant features, obscuring the attainment of accurate segmentations. To achieve semantic segmentation with functions of (1) extracting multi-scale features from images, (2) compensating information of high resolution, and (3) eliminating the task-irrelevant features, we propose a multi-scale model with skip connection framework and attention mechanism integrated. The multi-scale feature extraction modules are incorporated with additive attention gate units for irrelevant feature elimination, through a U-Net framework with skip connections for information compensation. The performance of fetal heart and lung segmentation indicates the superiority of our method over the existing deep learning based approaches. Our method also shows competitive performance stability during the task of semantic segmentations, showing a promising contribution on ultrasound based prognosis of congenital anomaly in the early intervention, and alleviating the negative effects caused by congenital anomaly.

Download Full-text

Multi-Scale Analysis of the UHVDC Transmission Tower

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.680.383 ◽

2014 ◽

Vol 680 ◽

pp. 383-386

Author(s):

Chun Cheng Liu ◽

Wen Qiang Li ◽

Shang Yu Hou ◽

Zhao Wen He ◽

Fan Gao

Keyword(s):

Large Scale ◽

Element Model ◽

Beam Element ◽

Scale Model ◽

Scale Analysis ◽

Transmission Tower ◽

Gusset Plate ◽

Multi Scale ◽

Buckling Behavior ◽

Scale Models

In order to analyze the mechanical properties of UHVDC transmission tower joint accurately, a multi-scale finite element model of the transmission tower is established with the interface between solid element model and beam element model. The model is applied to the nonlinear analysis of a key joint in a test condition .The results show that the tower destruction is caused by buckling behavior of the cross bracing member and the multi-scale model can simulate the force state of gusset-plate and the connected members realistically, which is superior to traditional large scale models. The analysis coincides with the experiment well and provides references for the transmission tower design.

Download Full-text

Enet Semantic Segmentation Combined with Attention Mechanism

10.21203/rs.3.rs-425438/v1 ◽

2021 ◽

Author(s):

Wei Bai

Keyword(s):

Feature Fusion ◽

Receptive Fields ◽

Design Feature ◽

Semantic Segmentation ◽

Attention Mechanism ◽

Segmentation Algorithm ◽

Feature Maps ◽

Model Learning ◽

Simple Method ◽

Feature Map

Abstract Image semantic segmentation is one of the core tasks of computer vision. It is widely used in fields such as unmanned driving, medical image processing, geographic information systems and intelligent robots. Aiming at the problem that the existing semantic segmentation algorithm ignores the different channel and location features of the feature map and the simple method when the feature map is fused, this paper designs a semantic segmentation algorithm that combines the attention mechanism. Firstly, dilated convolution is used, and a smaller downsampling factor is used to maintain the resolution of the image and obtain the detailed information of the image. Secondly, the attention mechanism module is introduced to assign weights to different parts of the feature map, which reduces the accuracy loss. The design feature fusion module assigns weights to the feature maps of different receptive fields obtained by the two paths, and merges them together to obtain the final segmentation result. Finally, through experiments, it was verified on the Camvid, Cityscapes and PASCAL VOC2012 datasets. Mean intersection over union (MIoU) and mean pixel accuracy (MPA) are used as metrics. The method in this paper can make up for the loss of accuracy caused by downsampling while ensuring the receptive field and improving the resolution, which can better guide the model learning. And the proposed feature fusion module can better integrate the features of different receptive fields. Therefore, the proposed method can significantly improve the segmentation performance compared to the traditional method.

Download Full-text

Fine Damage Identification Method of Transmission Tower Based on Concurrent Multi-Scale Modeling

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.680.374 ◽

2014 ◽

Vol 680 ◽

pp. 374-378

Author(s):

Chun Cheng Liu ◽

Shang Yu Hou ◽

Wen Qiang Li ◽

Zhao Wen He

Keyword(s):

Large Scale ◽

Damage Identification ◽

Wavelet Packet ◽

Fatigue Cracks ◽

Damage Index ◽

Scale Model ◽

Transmission Tower ◽

Damage Level ◽

Multi Scale ◽

Identification Method

In order to study the damage problem caused by the transmission tower fatigue cracks and bolt pretightening force loss ,this paper proposes a transmission tower damage identification method based on concurrent multi-scale model, namely establish solid model on nodes of fatigue crack and bolt looseness based on large scale model., subdividing elements size. Take a practical engineering 500kV transmission towers as an example to establish a concurrent multi-scale models. This paper simulates 8 kinds of conditions including bolt pretightening force loss and angle steel crack, research shows that the sum of wavelet packet energy curvature difference can effectively identify minute damage, and then get the function relation between damage level and damage index with no noise interference, also this provides a theoretical basis for it as actual damage monitoring indicators index.

Download Full-text

Multi-Scale Numerical Model for Groundwater flow Simulation of a Karst Tunnel in Kunming,China

10.5194/egusphere-egu2020-15112 ◽

2020 ◽

Author(s):

Mo Xu ◽

Jihong Qi ◽

Yige Tang ◽

Xiao Li ◽

JIan Guo

Keyword(s):

Groundwater Flow ◽

Water Level ◽

Large Scale ◽

Early Stage ◽

Water Discharge ◽

Scale Model ◽

Small Scale ◽

Tunnel Construction ◽

Multi Scale ◽

Pipe Model

Due to the inhomogeneity of the carbonate rocks and discreteness of the karst water, delineation of the groundwater flow within karst area remains a challenging task as yet. Based on KunCheng tunnel of a water diversion project in KunMing, multi-scale groundwater flow models were set upto simulate the groundwater flow. Large scale model was used to obtain the boundary conditions and hydrogeological parameters, which were then assigned to the small scale model.The small scale model was generalized as an equivalent continuous medium, and two karst pipelines are established &#160;by module River. After then, &#160;the multi-scale numerical modelswere used to simulate the &#160;groundwater seepage field and predict the recovery of groundwater after tunnel construction. The main results and conclusions are as follows.(1)Black karst pipeline and white karst pipeline systems share one recharge source but have two independent discharge systems. The recharge source is the exposed karst rock in the northeast part of study area. Obstructed by aluminum clay rock of P1d, groundwater discharge is divided into two parts during the runoff process.(2)During the tunnel construction process, the water level at the exit of White karst pipeline reduced 9m in pipe model B1 while reduced 10m in the solution fissure model B2, both two models suggest that the tunnel construction will cause the drainage of White karst pipeline exit. The water level at the exit of black karst pipeline reduced 1m in pipe model B1 while reduced 4m in the solution fissure model B2.(3)In model B1, total water discharge during tunnel construction is 69876m3/d, in model B2 , the total water discharge is 95817 m3/d &#160;and &#160;is much larger than model B1 due to the quick groundwater transporting and exchange in karst pipeline..(4)After the tunnel construction, exits of two pipelines and observation well see the water level recovery because of the formation sealing . The recovery trend is relatively rapid in the early stage, and slow in the later stage. It takes 8.5 years and 10 years for the exits of black and white pipelines and observation wells to reach the original water level, respectively. During the recovery process, groundwater exchange form was changing from pipe supplying aquifer to aquifer supplying pipe, which made model B2 recovered faster than model B1 in early stage, and vice versa.Using large scale model combining with secondary scale model, and the module River to generalize karst pipeline can reflect the flow dynamic characteristics of karst pipeline effectively.

Download Full-text

Study on Multi-Scale Coupled Ecological Dispatching Model Based on the Decomposition-Coordination Principle

Water ◽

10.3390/w11071443 ◽

2019 ◽

Vol 11 (7) ◽

pp. 1443

Author(s):

Zhou ◽

Dong ◽

Wang ◽

Shi ◽

Gao ◽

...

Keyword(s):

Large Scale ◽

Scale Model ◽

Small Scale ◽

Adaptive Genetic Algorithm ◽

Hydrologic Alteration ◽

Multi Scale ◽

Large Scale Model ◽

Calculation Results ◽

Coordination Principle ◽

Small Scale Model

Studies on environmental flow have developed into a flow management strategy that includes flow magnitude, duration, frequency, and timing from a flat line minimum flow requirement. Furthermore, it has been suggested that the degree of hydrologic alteration be employed as an evaluation method of river ecological health. However, few studies have used it as an objective function of the deterministic reservoir optimal dispatching model. In this work, a multi-scale coupled ecological dispatching model was built, based on the decomposition-coordination principle, and considers multi-scale features of ecological water demand. It is composed of both small-scale model and large-scale model components. The small-scale model uses a daily scale and is formulated to minimize the degree of hydrologic alteration. The large-scale model uses a monthly scale and is formulated to minimize the uneven distribution of water resources. In order to avoid dimensionality, the decomposition coordination algorithm is utilized for the coordination among subsystems; and the adaptive genetic algorithm (AGA) is utilized for the solution of subsystems. The entire model—which is in effect a large, complex system—was divided into several subsystems by time and space. The subsystems, which include large-scale and small-scale subsystems, were correlated by coordinating variables. The lower reaches of the Yellow River were selected as the study area. The calculation results show that the degree of hydrologic alteration of small-scale ecological flow regimes and the daily stream flow can be obtained by the model. Furthermore, the model demonstrates the impact of considering the degree of hydrologic alteration on the reliability of water supply. Thus, we conclude that the operation rules extracted from the calculation results of the model contain more serviceable information than that provided by other models thus far. However, model optimization results were compared with results from the POF approach and current scheduling. The comparison shows that further reduction in hydrologic alteration is possible and there are still inherent limitations within the model that need to be resolved.

Download Full-text

Pyramid Attention Aggregation Network for Semantic Segmentation of Surgical Instruments

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v34i07.6850 ◽

2020 ◽

Vol 34 (07) ◽

pp. 11782-11790

Author(s):

Zhen-Liang Ni ◽

Gui-Bin Bian ◽

Guan-An Wang ◽

Xiao-Hu Zhou ◽

Zeng-Guang Hou ◽

...

Keyword(s):

Specular Reflection ◽

Critical Role ◽

Receptive Fields ◽

Computer Assisted Surgery ◽

Semantic Segmentation ◽

Surgical Instruments ◽

Computer Assisted ◽

Multi Scale ◽

Semantic Dependencies ◽

Scale Variation

Semantic segmentation of surgical instruments plays a critical role in computer-assisted surgery. However, specular reflection and scale variation of instruments are likely to occur in the surgical environment, undesirably altering visual features of instruments, such as color and shape. These issues make semantic segmentation of surgical instruments more challenging. In this paper, a novel network, Pyramid Attention Aggregation Network, is proposed to aggregate multi-scale attentive features for surgical instruments. It contains two critical modules: Double Attention Module and Pyramid Upsampling Module. Specifically, the Double Attention Module includes two attention blocks (i.e., position attention block and channel attention block), which model semantic dependencies between positions and channels by capturing joint semantic information and global contexts, respectively. The attentive features generated by the Double Attention Module can distinguish target regions, contributing to solving the specular reflection issue. Moreover, the Pyramid Upsampling Module extracts local details and global contexts by aggregating multi-scale attentive features. It learns the shape and size features of surgical instruments in different receptive fields and thus addresses the scale variation issue. The proposed network achieves state-of-the-art performance on various datasets. It achieves a new record of 97.10% mean IOU on Cata7. Besides, it comes first in the MICCAI EndoVis Challenge 2017 with 9.90% increase on mean IOU.

Download Full-text

Combined equivalent & multi-scale simulation method for 3-D seismic analysis of large-scale shield tunnel

Engineering Computations ◽

10.1108/ec-02-2012-0034 ◽

2014 ◽

Vol 31 (3) ◽

pp. 584-620 ◽

Cited By ~ 9

Author(s):

Weiwei Zhang ◽

Xianlong Jin ◽

Zhihao Yang

Keyword(s):

Large Scale ◽

Simulation Method ◽

Equivalent Model ◽

Scale Model ◽

Shield Tunnel ◽

Fe Model ◽

Explicit Integration ◽

Integration Algorithm ◽

Content Type ◽

Multi Scale

Purpose – The great magnitude differences between the integral tunnel and its structure details make it impossible to numerically model and analyze the global and local seismic behavior of large-scale shield tunnels using a unified spatial scale, even with the help of supercomputers. The paper aims to present a combined equivalent & multi-scale simulation method, by which the tunnel's major mechanical properties under seismic loads can be represented by the equivalent model, and the seismic responses of the interested details can be studied efficiently by the coupled multi-scale model. Design/methodology/approach – The nominal orthotropic material constants of the equivalent tunnel model are inversely determined by fitting the modal characteristics of the equivalent model with the corresponding segmental lining model. The critical sections are selected by comprehensive analyzing of the integral compression/extension and bending loads in the equivalent lining under the seismic shaking and the coupled multi-scale model containing the details of interest is solved by the mixed time explicit integration algorithm. Findings – The combined equivalent & multi-scale simulation method is an effective and efficient way for seismic analyses of large-scale tunnels. The response of each flexible joint is related to its polar location on the lining ring, and the mixed time integration method can speed-up the calculation process for hybrid FE model with great differences in element sizes. Originality/value – The orthotropic equivalent assumption is, to the best of the authors’ knowledge, for the first time, used in the 3D simulation of the shield tunnel lining, representing the rigidity discrepancies caused by the structural property.

Download Full-text

A Feature-Enhanced Anchor-Free Network for UAV Vehicle Detection

Remote Sensing ◽

10.3390/rs12172729 ◽

2020 ◽

Vol 12 (17) ◽

pp. 2729

Author(s):

Jianxiu Yang ◽

Xuemei Xie ◽

Guangming Shi ◽

Wenzhe Yang

Keyword(s):

Large Scale ◽

Receptive Fields ◽

Vehicle Detection ◽

Feature Representation ◽

Multi Scale ◽

Discriminative Feature ◽

Aerial Vehicle ◽

Free Network ◽

Uav Images ◽

Localization Information

Vehicle detection based on unmanned aerial vehicle (UAV) images is a challenging task. One reason is that the objects are small size, low-resolution, and large scale variations, resulting in weak feature representation. Another reason is the imbalance between positive and negative examples. In this paper, we propose a novel architecture for UAV vehicle detection to solve above problems. In detail, we use anchor-free mechanism to eliminate predefined anchors, which can reduce complicated computation and relieve the imbalance between positive and negative samples. Meanwhile, to enhance the features for vehicles, we design a multi-scale semantic enhancement block (MSEB) and an effective 49-layer backbone which is based on the DetNet59. The proposed network offers appropriate receptive fields that match the small-sized vehicles, and involves precise localization information provided by the contexts with high resolution. The MSEB strengthens discriminative feature representation at various scales, without reducing the spatial resolution of prediction layers. Experiments show that the proposed method achieves the state-of-the-art performance. Particularly, the main part of vehicles, much smaller ones, the accuracy is about 2% higher than other existing methods.

Download Full-text

A hybrid multi-scale model for thyroid nodule boundary detection on ultrasound images

Computer Methods and Programs in Biomedicine ◽

10.1016/j.cmpb.2006.09.006 ◽

2006 ◽

Vol 84 (2-3) ◽

pp. 86-98 ◽

Cited By ~ 29

Author(s):

S. Tsantis ◽

N. Dimitropoulos ◽

D. Cavouras ◽

G. Nikiforidis

Keyword(s):

Thyroid Nodule ◽

Boundary Detection ◽

Scale Model ◽

Ultrasound Images ◽

Multi Scale

Download Full-text

Multi-Level and Multi-Scale Feature Aggregation Network for Semantic Segmentation in Vehicle-Mounted Scenes

Sensors ◽

10.3390/s21093270 ◽

2021 ◽

Vol 21 (9) ◽

pp. 3270

Author(s):

Yong Liao ◽

Qiong Liu

Keyword(s):

Receptive Fields ◽

Semantic Segmentation ◽

Layer By Layer ◽

Scale Feature ◽

Backbone Networks ◽

Multi Scale ◽

Backbone Network ◽

Feature Aggregation ◽

Multi Level ◽

High Level

The main challenges of semantic segmentation in vehicle-mounted scenes are object scale variation and trading off model accuracy and efficiency. Lightweight backbone networks for semantic segmentation usually extract single-scale features layer-by-layer only by using a fixed receptive field. Most modern real-time semantic segmentation networks heavily compromise spatial details when encoding semantics, and sacrifice accuracy for speed. Many improving strategies adopt dilated convolution and add a sub-network, in which either intensive computation or redundant parameters are brought. We propose a multi-level and multi-scale feature aggregation network (MMFANet). A spatial pyramid module is designed by cascading dilated convolutions with different receptive fields to extract multi-scale features layer-by-layer. Subseqently, a lightweight backbone network is built by reducing the feature channel capacity of the module. To improve the accuracy of our network, we design two additional modules to separately capture spatial details and high-level semantics from the backbone network without significantly increasing the computation cost. Comprehensive experimental results show that our model achieves 79.3% MIoU on the Cityscapes test dataset at a speed of 58.5 FPS, and it is more accurate than SwiftNet (75.5% MIoU). Furthermore, the number of parameters of our model is at least 53.38% less than that of other state-of-the-art models.

Download Full-text