Search | Engineering

订阅投稿

首页工程期刊工程焦点工程成就工程前沿关于我们 English

资源类型

期刊论文 564

会议视频 17

会议信息 11

会议专题 1

年份

2024 1

2023 84

2022 82

2021 67

2020 55

2019 52

2018 37

2017 48

2016 16

2015 15

2014 4

2013 7

2012 9

2011 10

2010 11

2009 8

2008 11

2007 16

2006 5

2005 7

展开︾

关键词

机器学习 27

深度学习 15

人工智能 14

2019 3

MATLAB 3

图像处理 3

2020 2

3D打印 2

ACM 2

一阶分析法 2

乳液 2

仿真 2

分布式优化 2

区块链 2

增材制造 2

强化学习 2

材料设计 2

海上风电场 2

结构健康监测 2

展开︾

检索范围：

排序：展示方式：

结合全变分最小化和稀疏字典学习后处理的低剂量CT重建 Article

Yong DING, Tuo HU

《信息与电子工程前沿（英文）》 2017年第18卷第12期页码 2001-2008 doi: 10.1631/FITEE.1700287

摘要：随着CT（computed tomography）中过量辐射剂量带来的健康风险日渐引发人们的担忧，低剂量CT得到了大量的关注。然而对于低剂量CT成像而言，在降低剂量的同时保证图像的高质量是一个很大的挑战。相比传统的滤波反投影算法，基于压缩感知的迭代重建法取得了良好的成像效果。但是迭代重建计算复杂度高，阻碍了其临床应用。本文提出一种结合全变分（total variation, TV）最小化和稀疏字典学习的重建方法，不仅提高了重建效果，而且通过自适应的停止策略提高了重建速度。

关键词：低剂量CT；CT成像；全变分；稀疏字典学习

HTML PDF 收藏

基于稀疏表示的拉普拉斯稀疏字典图像分类 Article

Fang LI, Jia SHENG, San-yuan ZHANG

《信息与电子工程前沿（英文）》 2017年第18卷第11期页码 1795-1805 doi: 10.1631/FITEE.1600039

摘要：稀疏表示作为数据表示的一种数学模型，是解决模式识别、机器学习、计算机视觉等领域问题的有力工具。字典学习是稀疏表示方法的重要组成部分，在对原始信号及其在字典学习空间中的重建误差的最小化上发挥着重要的作用。在稀疏表示模型中，直接利用训练样本作为字典可以取得良好的性能。但由于训练样本含有噪声，这样的字典很大且效率低下。为取得更小且表现更好的字典，本文提出一种基于流形学习及双稀疏理论的拉普拉斯稀疏字典学习方法（Laplacian sparse dictionary, LSD）。本文将拉普拉斯权重图加入稀疏表示的模型，并对字典加以范数约束。LSD是一个稀疏的过完备字典，可保持数据的内在结构，并为每个类学习一个更小的字典。学习得到的字典可以嵌入基于稀疏表示的分类框架。

关键词：稀疏表示；拉普拉斯正则子；字典学习；双稀疏；流形

HTML PDF 收藏

Simultaneous acquisition of CT angiography and whole brain CT perfusion images by using multiphase dynamic

Weiwei CHEN, Jianpin QI, Wenzhen ZHU, Wenhua HUANG, Jinmei SONG

《医学前沿（英文）》 2009年第3卷第2期页码 230-235 doi: 10.1007/s11684-009-0023-8

摘要： We implemented a new protocol — multiphase dynamic helical scan to acquire CT angiography (CTA) and whole brain CT perfusion (CTP) images simultaneously with single scan on 16 multidetector CT (MDCT). A total of 90 patients who were randomly assigned into 3 groups were included in our study. Each group underwent CT scan by using the new protocol, traditional CTA and CTP protocol, respectively. The image quality of CTA, the CTP parameter values and the X-ray doses were measured and compared between the new protocol and the traditional protocols. There was no statistically significant difference in the CTA image quality between the above methods ( =0.55). For CTP parameters, the new protocol tended to overestimate the blood volume (BV) and blood flow (BF) value, and to underestimate the mean transit time (MTT) value compared with the traditional method. However, there was no statistically significant difference in BV, BF, and MTT value between the two methods except permeability surface (PS) ( >0.05). The volume CT dose index (CTDIvol) and dose length product (DLP) of our protocol were lower than the traditional one. The new protocol can obtain valuable diagnostic information in a shorter time without significant compromise in image quality. In addition, it reduces the radiation dose as well as contrast medium usage on the patient.

关键词： computed tomography angiography computed tomography perfusion

HTML PDF 收藏

Low-dose CT for lung cancer screening: opportunities and challenges

null

《医学前沿（英文）》 2018年第12卷第1期页码 116-121 doi: 10.1007/s11684-017-0600-1

摘要：

Lung cancer is among the most frequently diagnosed cancers worldwide and the leading cause of cancer death in both males and females. Screening for lung cancer coupled with earlier intervention has long been studied as an approach to mortality reduction. However, minimal progress was achieved until recently, when low-dose spiral computed tomography (LDCT) screening demonstrated a 20% reduction in mortality from lung cancer in a randomized controlled trial (RCT), the National Lung Screening Trial, from the United States. On the basis of this finding, LDCT has been recommended for lung cancer screening in high-risk populations by several clinical guidelines. However, results from the following independent RCTs in Europe failed to show consistent conclusions. In addition, intractable problems gradually emerged with the progress of LDCT screening. This paper summarizes and discusses the main observations and challenges of LDCT screening for lung cancer. Before spreading implementation of LDCT screening, challenges, including high false-positive rates, overdiagnosis, enormous costs, and radiation risk, must be addressed. Complementary biomarkers and technical improvement are expected in the field of lung cancer screening in the near future.

关键词： lung cancer low-dose computerized tomography early detection opportunities challenges

HTML PDF 收藏

The dynamic correction of collimation errors of CT slicing pictures

LIU Ya-xiong, Sekou Sing-are, LI Di-chen, LU Bing-heng

《机械工程前沿（英文）》 2006年第1卷第2期页码 168-172 doi: 10.1007/s11465-006-0016-0

摘要： To eliminate the motion artifacts of CT images caused by patient motions and other related errors, two kinds of correctors (A type and U type) are proposed to monitor the scanning process and correct the motion artifacts of the original images via reverse geometrical transformation such as reverse scaling, moving, rotating and offsetting. The results confirm that the correction method with any of the correctors can improve the accuracy and reliability of CT images, which facilitates in eliminating or decreasing the motion artifacts and correcting other static errors and image processing errors. This provides a foundation for the 3D reconstruction and accurate fabrication of the customized implants.

HTML PDF 收藏

基于核稀疏表示的磁共振图像分析及其在脑肿瘤自动分割中的应用 None

Ji-jun TONG, Peng ZHANG, Yu-xiang WENG, Dan-hua ZHU

《信息与电子工程前沿（英文）》 2018年第19卷第4期页码 471-480 doi: 10.1631/FITEE.1620342

摘要：提出一种基于核稀疏编码的全自动脑肿瘤分割方法，并在3D多模态磁共振成像图（magnetic resonance imaging, MRI）上验证。首先对MRI图像进行预处理以减少噪声，然后通过核字典学习提取非线性特征，用来构建坏死组织、水肿组织、非增强肿瘤组织、增强肿瘤组织和健康组织5个适应性字典。对从原始MRI图像上肿瘤像素点周边m×m×m的小区域提取的特征向量进行稀疏编码，并通过一种基于字典学习的核聚类方法对像素点进行编码。最后通过形态滤波填充在多个相连部分间的区域，提高分割质量。

关键词：脑肿瘤分割；核方法；稀疏编码；字典学习

HTML PDF 收藏

一种用于工业过程监测的鲁棒迁移字典学习算法 Article

阳春华, 梁慧平, 黄科科, 李勇刚, 桂卫华

《工程（英文）》 2021年第7卷第9期页码 1262-1273 doi: 10.1016/j.eng.2020.08.028

摘要：因此，当将从训练数据中学习的模型应用于实际的在线监测时，很难实现精确的过程监测。为了解决操作环境变化导致的历史训练数据和在线测试数据之间的分布差异问题，提出了一种鲁棒的迁移字典学习（RTDL）算法用于工业过程监测。RTDL是表示学习和域自适应迁移学习的协同方法。该方法将历史训练数据和在线测试数据分别作为迁移学习问题的源域和目标域。然后将最大均值差异正则化和线性判别分析正则化引入字典学习框架，可以减少源域和目标域之间的分布差异。这样，即使源域和目标域的特征在实际变化的操作环境的干扰下明显不同，仍可以学习鲁棒的字典。这样的字典可以有效地提高过程监测和模态识别的性能。通过数值仿真和两个工业系统的实验验证了该方法的有效性和优越性。

关键词：过程监测多模态过程字典学习迁移学习

HTML PDF 收藏

Analyzing the characterization of pore structures and permeability of diesel contaminated clays under different aging conditions

《结构与土木工程前沿（英文）》页码 1264-1280 doi: 10.1007/s11709-023-0921-x

摘要： In this study, mercury intrusion porosimetry (MIP) and X-ray micro-computed tomography (XRμCT) were used to characterize the pore structures and investigate the permeability characteristics of clay after aging and contamination with diesel. The results of the MIP tests showed that aging leads to reductions in porosity and average diameter, as well as an increase in tortuosity. The XRμCT analysis yielded consistent results; it showed that aging renders pores more spherical and isotropic and pore surfaces smoother. This weakens the pore connectivity. Micromorphological analysis revealed that aging led to the rearrangement of soil particles, tighter interparticle overlapping, and a reduction in pore space. The combination of MIP and XRμCT provided a comprehensive and reliable characterization of the soil pore structure. An increased diesel content increased the porosity and average diameter and reduced the tortuosity of the pores. Mechanistic analysis showed that aging weakens interparticle cohesion; this causes large agglomerates to break down into smaller agglomerates, resulting in a tighter arrangement and a subsequent reduction in porosity. An increase in diesel content increases the number of large agglomerates and pore spaces between agglomerates, resulting in increased porosity. Both aging and diesel content can weaken the permeation characteristics of soil.

关键词： MIP XRμCT aging diesel content pore structure permeability characteristics

HTML PDF 收藏

Creep-fatigue crack growth behavior in GH4169 superalloy

Dianyin HU, Xiyuan WANG, Jianxing MAO, Rongqiao WANG

《机械工程前沿（英文）》 2019年第14卷第3期页码 369-376 doi: 10.1007/s11465-018-0489-7

摘要： This study aims to examine the crack growth behavior of turbine disc GH4169 superalloy under creep-fatigue loading. Crack growth experiments were performed on compact tension specimens using trapezoidal waveform with dwell time at the maximum load at 650 °C. The crack growth rate of GH4169 superalloy significantly increased with dwell time. The grain boundaries oxidize during the dwell process, thereby inducing an intergranular creep-fatigue fracture mode. In addition, testing data under the same dwell time showed scattering at the crack growth rate. Consequently, a modified model based on the Saxena equation was proposed by introducing a distribution factor for the crack growth rate. Microstructural observation confirmed that the small grain size and high volume fraction of the d phase led to a fast creep-fatigue crack growth rate at 650 °C, thus indicating that two factors, namely, fine grain and presence of the d phase at the grain boundary, increased the amount of weakened interface at high temperature, in which intergranular cracks may form and propagate.

关键词： crack growth rate creep-fatigue GH4169 superalloy CT specimen dwell time

HTML PDF 收藏

A pilot study on Paxlovid therapy for hemodialysis patients with severe acute respiratory syndrome coronavirus 2 infections

《医学前沿（英文）》 doi: 10.1007/s11684-023-1011-0

摘要： We aimed to investigate the safety and efficacy of nirmatrelvir/ritonavir (Paxlovid) therapy for hemodialysis-dependent patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Thirteen hemodialysis patients infected with the Omicron variant of SARS-CoV-2 from April 3 to May 30, 2022, were recruited. Laboratory parameters and chest CT (computed tomography) imaging were analyzed. The treatment group included six patients who received 150 mg/100 mg of Paxlovid orally once daily for 5 days, whereas the control group included seven patients who received basic treatment. No serious adverse reactions or safety events were recorded. Four control patients progressed to moderate disease, and none in the treatment group showed progression of chest CT findings (P < 0.05). Paxlovid therapy tended toward early viral clearance and low viral load on Day 8. Moreover, 83.3% of the patients in the treatment group and 57.1% of the patients in the control group turned negative within 22 days. In the Paxlovid treatment group, we found significantly increased levels of lymphocytes (P=0.03) and eosinophils (P=0.02) and decreased levels of D-dimer on Day 8 compared with those on Day 1. Paxlovid therapy showed a potential therapeutic effect with good tolerance in hemodialysis patients. The optimal dose and effectiveness evaluation must be further investigated in a largeer cohort.

关键词： Paxlovid hemodialysis SARS-CoV-2 viral load chest CT scan

HTML PDF 收藏

Comparison of CT-guided aspiration to key hole craniotomy in the surgical treatment of spontaneous putaminal

ZHAO Jizong, WANG Dejiang, WANG Shuo, YUAN Ge, KANG Shuai, JI Nan, ZHAO Yuanli, YE Xun, ZHOU Liangfu, ZHOU Dingbiao, WANG Renzhi, WANG Mei

《医学前沿（英文）》 2007年第1卷第2期页码 142-146 doi: 10.1007/s11684-007-0027-1

摘要： This study was designed to compare the approaches and efficacies of two different ways of neurosurgical management for spontaneous putaminal hemorrhage (SPH): computed tomographic-guided aspiration (CTGA) and the key-hole approach (KHA). The indications of the two approaches are also explored. From September 2001 to 2003, a total of 1077 cases of SPH distributed in 135 hospitals all over the mainland of China were included for analysis. All cases had three-month follow-up data. The study was designed in a single-blinded manner to compare the efficacies of the different approaches. There were 563 cases in the CTGA group, 165 in the KHA group, and 217 cases in the conventional open craniotomy (COC) group. In the CTGA and KHA groups, the mortalities at one month after operation (M1m) were 17.9% and 18.3%, respectively, while the mortalities at three months after operation (M3m) were 19.4% and 19.4%, respectively (>0.05). The postoperative complications due to CTGA (23.7%) were not significantly different from those due to KHA (25.7%) ( = 0.420). The M3m of patients with Glasgow coma scale (GCS) "d8 was 3.45 and 4.0 times as much as those with GCS>8, respectively. The M3m of patients with complications was 3.92 times as much as those without complications. The M3m of patients with hemorrhage volume "e70 mL was 2.67 times as much as those <70 mL. The CTGA is not better than KHA in the treatment of SPH in terms of a more favorable outcome or less mor tality and morbidity, but CTGA could be the first choice for those with bleeding volumes "d50mL, while KHA is the first choice for those with bleeding volumes >50 mL.

关键词： tomographic-guided aspiration single-blinded manner favorable outcome operation Glasgow

HTML PDF 收藏

结构化稀疏学习综述 Review

Lin-bo QIAO, Bo-feng ZHANG, Jin-shu SU, Xi-cheng LU

《信息与电子工程前沿（英文）》 2017年第18卷第4期页码 445-463 doi: 10.1631/FITEE.1601489

摘要：稀疏学习由于其简约特性和计算优势而获得了越来越多的关注，在具有稀疏性的条件下，许多计算问题可以在实践中得到有效的处理。而结构化稀疏学习则进一步将结构信息进行编码，在多个研究领域取得成功。这些正则函数通过利用特定的结构信息极大提高了稀疏学习算法的性能。在本文中，我们从想法、形式化、算法和应用等方面系统的回顾了结构化稀疏学习。我们将这些算法置于最小化损失函数和惩罚函数的统一框架中，总结了算法的开源软件实现，并比较了典型优化算法解决结构化稀疏学习问题时的计算复杂度。在实验中，我们给出了无监督学习在结构化信号恢复和层次化图像重建中的应用，以及具有图结构引导的逻辑回归的在监督学习中的应用。

关键词：结构化稀疏学习；算法；应用

HTML PDF 收藏

基于实时CT扫描技术的CO₂和N₂交替注入条件下煤吸附膨胀和解吸收缩规律研究 Article

张广磊, P.G. Ranjith, Herbert E. Huppert

《工程（英文）》 2022年第18卷第11期页码 88-95 doi: 10.1016/j.eng.2022.03.010

摘要：本文采用原位同步辐射X射线显微CT扫描技术，首次在原位条件下直接证明了注入氮气（N₂）可以置换解吸CO₂并减小因CO₂吸附引起的煤基质膨胀

关键词： CCS CO₂-ECBM 碳中和 X 射线成像煤渗透率

HTML PDF 收藏

基于变分贝叶斯多稀疏成分提取的空间碎片超高速撞击损伤重构方法研究 Research Article

黄雪刚,石安华,罗庆,罗锦阳

《信息与电子工程前沿（英文）》 2022年第23卷第4期页码 530-541 doi: 10.1631/FITEE.2000575

摘要：采用变分贝叶斯推理来求解模型参数，从而有效地从红外热图像数据中识别不同类型撞击损伤。然后，提出一种图像处理框架，包括具有能量函数的图像分割算法和具有稀疏表示的图像融合方法，以消除变异贝叶斯误差并比较不同类型损伤的位置。

关键词：超高速撞击；变分贝叶斯；稀疏表示；损伤评估

HTML PDF 收藏

联合局部学习和组稀疏回归的无监督特征选择 Regular Papers

Yue WU, Can WANG, Yue-qing ZHANG, Jia-jun BU

《信息与电子工程前沿（英文）》 2019年第20卷第4期页码 538-553 doi: 10.1631/FITEE.1700804

摘要：通过挑选特征子集，可有效提升学习算法效率。由于难以获取标签信息，无监督特征选择算法相较于有监督特征选择算法应用更为广泛，其关键在于找出更能反映数据分布的特征集合。为解决这一问题，本文提出联合局部学习和组稀疏回归的无监督特征选择算法。将基于局部学习聚类方法与组稀疏回归算法有机整合，选出有效反映数据流形分布同时保持组稀疏结构的特征。

关键词：无监督；局部学习；组稀疏回归；特征选择