摘要：
U-network is a comprehensive convolutional neural network that is widely utilized in medical image segmentation domain. However, it is not accurate enough in detail segmentation and resulting in unsatisfactory segmentation results. To solve this problem, this paper proposes an enhanced U-network that combines an improved Pyramid Pooling Module (PPM) and a modified Convolutional Block Attention Module (CBAM). Its whole network is U-Net architecture, where the PPM is improved by reducing the number of bin species and increasing the pooling connection multiples. It is used in the downsampling part of the network, which can extract input image features of various dimensions. And the CBAM is modified by using 1x1 convolutional layers instead of the original fully connected layers. It is used in the upsampling part of the network, which can combine convolution and attention mechanism. This pays attention to the image from two aspects of space and channel. Besides, the network is trained with novel RGB training to further improve the segmentation ability of the network. Experimental results show that our network outperforms traditional U-shaped segmentation networks by 30% to 40% in metrics Dice, IoU, MAE, and BFscore respectively. What's more, it is better than U-Net ++, U2-Net, ResU-Net, ResU-Net++, and UNeXt in terms of segmentation effect and training time.
期刊：
Journal of Biomedical Informatics,2019年91:103114 ISSN：1532-0464
通讯作者：
Wang, Jianxin
作者机构：
[Fei, Zhihui; Li, Min; Yu, Ying; Liu, Liangliang; Wang, Jianxin] Cent South Univ, Sch Comp Sci & Engn, Changsha 410083, Peoples R China.;[Yu, Ying] Univ South China, Sch Comp Sci & Technol, Hengyang 421001, Peoples R China.;[Wu, Fang-Xiang] Univ Saskatchewan, Div Biomed Engn, Saskatoon, SK S7N 5A9, Canada.;[Wu, Fang-Xiang] Univ Saskatchewan, Dept Mech Engn, Saskatoon, SK S7N 5A9, Canada.
通讯机构：
[Wang, Jianxin] C;Cent South Univ, Sch Comp Sci & Engn, Changsha 410083, Peoples R China.
关键词：
Codes (symbols);Embeddings;Multilayers;Semantics;Bidirectional recurrent neural networks;Character-enhanced;Clinical notes;Electronic health record;Hierarchical approach;ICD code;International classification of disease;Neural network model;Recurrent neural networks;Article;artificial neural network;attention;Chinese (language);hospital admission;human;International Classification of Diseases;linguistics;machine learning;medical record;multilayer attention bidirectional recurrent neural network;performance;prediction;priority journal;problem solving;short term memory;automation;China;electronic health record;information processing;machine learning;Automation;China;Datasets as Topic;Electronic Health Records;International Classification of Diseases;Machine Learning
摘要：
International Classification of Diseases (ICD) code is an important label of electronic health record. The automatic ICD code assignment based on the narrative of clinical documents is an essential task which has drawn much attention recently. When Chinese clinical notes are the input corpus, the nature of Chinese brings some issues that need to be considered, such as the accuracy of word segmentation and the representation of single Chinese characters which contain semantics. Taking the lengthy text of patient notes and the representation of Chinese words into account, we present a multilayer attention bidirectional recurrent neural network (MA-BiRNN) model to implement the assignment of disease codes. A hierarchical approach is used to represent the feature of discharge summaries without manual feature engineering. The combination of character level embedding and word level embedding can improve the representation of words. Attention mechanism is introduced into bidirectional long short term memory networks, which helps to solve the performance dropping problem when plain recurrent neural networks encounter long text sequences. The experiment is carried out on a real-world dataset containing 7732 admission records in Chinese and 1177 unique ICD-10 labels. The proposed model achieves 0.639 and 0.766 in F1-score on full-level code and block-level code, respectively. It outperforms the baseline neural network models and achieves the lowest Hamming loss value. Ablation analysis indicates that the multilevel attention mechanism plays a decisive role in the system for dealing with Chinese clinical notes.
作者：
Ying Yu;Min Li 0007;Liangliang Liu 0001;Yaohang Li;Jianxin Wang 0001
期刊：
大数据挖掘与分析(英文),2019年2(4):288-305 ISSN：2096-0654
通讯作者：
Wang, Jianxin(yaohang@odu.edu)
作者机构：
[Ying Yu; Min Li 0007; Liangliang Liu 0001; Jianxin Wang 0001] School of Computer Science and Engineering, Central South University, Changsha;410083, China;School of Computer Science and Technology, University of South China, Hengyang;421001, China;[Yaohang Li] Department of Computer Science, Old Dominion University, Norfolk
通讯机构：
[Wang, J.] S;School of Computer Science and Engineering, China
关键词：
Clinical data;Clinical note;Deep learning;Electronic health record (EHR);Medical image
作者机构：
[刘立; 罗扬; 汪琳霞; 刘芳菊; 李悛] School of Computer Science and Technology, University of South China, Hengyang;Hunan;421001, China;[刘立; 罗扬; 汪琳霞; 刘芳菊; 李悛] Hunan;[刘立; 罗扬; 汪琳霞; 刘芳菊; 李悛] 421001, China
通讯机构：
[Luo, Y.] S;School of Computer Science and Technology, China
期刊：
Proceedings of the 2015 27th Chinese Control and Decision Conference, CCDC 2015,2015年:1975-1979 ISSN：1948-9439
通讯作者：
Wu, Daqing
作者机构：
[Wu, Daqing] Anhui Univ, Minist Educ, Key Lab Intelligent Comp & Signal Proc, Hefei 230039, Anhui, Peoples R China.;[Liu, Li; Gong, XiangJian; Wu, Daqing] Univ South China, Comp Sci & Technol Inst, Hengyang 421001, Hunan, Peoples R China.;[Deng, Li; Wu, Daqing] DongHua Univ, Glorious Sun Sch Business & Management, Shanghai 200051, Peoples R China.
通讯机构：
[Wu, Daqing] A;Anhui Univ, Minist Educ, Key Lab Intelligent Comp & Signal Proc, Hefei 230039, Anhui, Peoples R China.
会议名称：
The 27th Chinese Control and Decision Conference (2015 CCDC)
会议时间：
May 2015
会议地点：
Qingdao, China
会议主办单位：
[Wu, Daqing] Anhui Univ, Minist Educ, Key Lab Intelligent Comp & Signal Proc, Hefei 230039, Anhui, Peoples R China.^[Wu, Daqing;Liu, Li;Gong, XiangJian] Univ South China, Comp Sci & Technol Inst, Hengyang 421001, Hunan, Peoples R China.^[Wu, Daqing;Deng, Li] DongHua Univ, Glorious Sun Sch Business & Management, Shanghai 200051, Peoples R China.
会议论文集名称：
The 27th Chinese Control and Decision Conference (2015 CCDC)
关键词：
Multi-objective Optimization;Particle Swarm Optimizer;Neighborhood Best Particle;Dynamic Swarms;Economic Environmental Dispatch
摘要：
An efficient co-evolutionary multi-objective particle swarm optimizer named ECMPSO was proposed. ECMPSO uses dynamic multiple swarms to deal with multiple objectives, taking one objective is optimized by each swarm into account, and maintains diversity of new found non-dominated solutions via adopts a three-level particle swarm optimization(PSO) updating rule wherein the particles learn their experiences based on personal, neighborhood, and external archive. To prove the validity of the ECMPSO algorithm for solving multi-objective problems, some benchmark problems and one real-life problem are selected to validate the performance of the ECMPSO algorithm. The experiment results show that the ECMPSO algorithm is better in terms of search precision and convergence performance than other three algorithms from the literature.
作者机构：
[刘立; 伍大清] Computer Science and Technology Institute, University of South China, Hengyang, China;[伍大清] Key Laboratory of Agri-informatics, Ministry of Agriculture, Beijing, China;[伍大清] Artificial Intelligence Key Laboratory of Sichuan Province, Sichuan University of Science and Engineering, Zigong, China;[郑建国; 朱君璇; 伍大清; 赵燕] School of Business and Management, Donghua University, Shanghai, China;[伍大清] Key Laboratory of Pattern Recognition and Intelligent Information Processing of Sichuan Province, Chengdu University, Chengdu, China
通讯机构：
[Huang, XinYang] U;Univ S China, Sch Comp Sci & Technol, Hengyang, Peoples R China.
会议名称：
International Workshop on Information and Electronics Engineering (IWIEE) / International Conference on Information, Computing and Telecommunications (ICICT)
会议时间：
MAR 10-11, 2012
会议地点：
Harbin, PEOPLES R CHINA
会议主办单位：
[Huang, XinYang;Tan, MinSheng;Ouyang, Lijun;Liu, Li] Univ S China, Sch Comp Sci & Technol, Hengyang, Peoples R China.
会议论文集名称：
Procedia Engineering
关键词：
DWT;Xiong invariant wavelet(XIW);rotation- and scaling- and translation-(RST) invariance;crop- invariance
摘要：
This paper presents a new method to solve the crop- invariant problem of an invariant wavelet, RSTXIW, which is a rotation- and scaling- and translation- invariant. Based on the feature points and second moments of the signal and based on a scale function of Daubiechies wavelet, we renormalize a signal in the function space. By using the feature points to divide the signal into several parts and by applying cluster analysis, we can correctly calculate the whole second-order moments of the cropped signal. Then we can obtain the crop- invariance of the renormalized signal, tested by some experiments. (C) 2011 Published by Elsevier Ltd.
